Cybernetics, Cognition and Machine Learning Applications: Proceedings of ICCCMLA 2021 (Algorithms for Intelligent Systems) 9811914834, 9789811914836

Table of contents :
Preface
Contents
About the Editors
A Deep Learning Approach to Analyze Traffic Congestions for Effective Traffic Management
1 Introduction
2 Existing Features
3 Proposed System
4 Methodology
5 Results
6 Conclusion and Future Work
References
Building Energy Consumption Prediction Model Using Machine Learning
1 Introduction
2 Literature Review
3 Background
3.1 Artificial Intelligence
3.2 Machine Learning
3.3 Linear Regression
4 Methodology
5 Results
6 Conclusion
References
An End-to-End GUI-Based Real-Time Attendance System by Training Annotated Facial Data on YOLO v3 Algorithm
1 Introduction
2 Literature Review
3 Methodology
4 Materials and Methods
5 Results
6 Conclusion
References
A Multifactor Security Protocol for Wireless Payment-Secure Web Authentication Using Mobile Devices
1 Introduction
1.1 Types of Authentication Systems
2 Literature Survey
3 Existing Protocol/System
4 Designed System
4.1 Protocol
5 Security
6 Conclusion
References
IoT-Based Greenhouse Monitoring
1 Introduction
2 Literature Review
3 Design of Greenhouse Monitoring
3.1 Block Diagram
3.2 Hardware Description
3.3 Software Description
4 Methodology
5 Result
6 Conclusion
References
Home Automation Using Telegram Bot
1 Introduction
2 Literature Survey
2.1 Chat Box
2.2 Run Your Own Bot API Server
3 Design Methodology
3.1 Architecture
4 Implementation
4.1 Telegram Bot
4.2 Raspberry Pi Camera
4.3 DHT11
4.4 Two-channel Relay
4.5 DC Fan
5 Software Tools
5.1 Telepot and Data Publishing in Telegram App
6 Results and Discussions
7 Conclusion
References
Application of IoT in Hospital Management
1 Introduction
1.1 Internet of Things
1.2 Vascular Air Embolism
1.3 Fire Accident
1.4 Power Consumption
2 Literature Survey
3 Block Diagram
3.1 Components
4 Design and Functionality
4.1 Saline Level Indicator
4.2 Harmful Gas Detection
4.3 Reduction of Power Consumption
5 Hardware Connections
6 Results and Discussion
7 Conclusion
References
Vehicle Tracking by Using IoT Applications
1 Introduction
2 Methodology
3 Results and Discussion
4 Conclusion
References
Smart Vehicle and Smart Parking System Using IOT
1 Introduction
2 Methodology
3 About Blynk App
4 Results and Discussion
5 Conclusion
References
Automatic Water Irrigation System Using IoT
1 Introduction
2 Proposed System
3 Results
4 Conclusion
References
Development of Safety Monitoring for an IOT-Enabled Smart Environment
1 Introduction
1.1 Video Processing
1.2 OpenCV & NumPy
2 Related Work
3 Safety Monitoring for Smart Environment System Overview
3.1 Ultrasonic Detector
3.2 Rain Detector
3.3 GSM
4 Triangulation Method for Camera Arrangement
5 Conclusion
References
Low-Cost ECG-Based Heart Monitoring System with Ubidots Platform
1 Introduction
2 System Architecture
3 Methodology
4 Software Implementation
5 Experimental Measurements and Results
6 Cost Analysis
7 Conclusion
References
Design of Hybrid Soft Computing Techniques for Estimation of Suspended Sediment Yield in Krishna River, India
1 Introduction
2 Study Area and Data Used
3 Methodology
4 Result and Discussion
5 Conclusions
References
QR-Based Ticket Verification and Parking System
1 Introduction
2 Literature Survey
3 Related Work
4 Proposed Work
5 Implementation Model
5.1 QR Code Ticket Generation and Scanning
5.2 QR Code for Parking Stand
6 Block Diagram for Proposed Model
6.1 QR Code Verification
6.2 Flow Chart for Parking Allotment and Payment
6.3 Pseudo Code for Ticket Verification
6.4 Pseudo Code for Parking System
7 Experimental Results
8 Conclusion
References
Design and Analysis of Uniformly Illuminated 8-Way Wilkinson Power Divider for L-Band Applications
1 Introduction
2 The Wilkinson Power Divider with Numerical Illustrations
2.1 Effective Dielectric Constant
2.2 Characteristic Impedance
2.3 Width of the Microstrip
2.4 Length of the Microstrip
3 The Proposed Design and Analysis
4 Results and Discussions
4.1 S-Parameters
4.2 Electric Field (E-field)
4.3 Magnetic Field (H-Field)
4.4 Surface Current
5 Conclusion
References
Comparative Analysis on Mulberry Leaf Disease Detection Using SVM and PNN
1 Introduction
2 Literature Survey
3 Plant Disease Identification
4 Results and Discussion
5 Conclusion
References
Thermal Analysis of Different Components on the PCB Using ANSYS Software
1 Introduction
2 Thermal Analysis of PCBs
3 ANSYS Tool for Thermal Heat Simulation
4 Simulation and Results
5 Conclusion
References
Custom Handloom Mobile Application
1 Introduction
2 Literature Survey
3 Requirements
4 Proposed System
5 Results
6 Conclusion
References
A Compact Orthomode Transducer in K-Band for Satellite Communications
1 Introduction
2 OMT Design
3 Simulated Results and Discussions
4 Conclusion
References
A Review on Stock Market Analysis Using Association Rule Mining
1 Introduction
2 Literature Survey
3 Association Rule on Stock Market
4 Mining Various Kinds of Knowledge from Databases
5 The Methodology of the Study
6 Conclusion
References
Student Performance Assessment Using AI
1 Introduction
2 Literature Survey
3 Models
4 Investigation
5 Conclusion
References
Automatic Smoke Absorber and Filter
1 Introduction
2 Literature Review
2.1 Problem Statement
2.2 Objectives
2.3 Existing Solutions
3 Project Design
3.1 Requirement Analysis
3.2 Implementation
3.3 Block Diagram
3.4 Applications
4 Conclusion
References
Computational Intelligence in Subthalamic Nucleus Deep Brain Stimulation: Machine Learning Unsupervised PCA Tracking Method and Clustering Techniques for Parkinson’s Feature Extraction
1 Introduction
2 Hypothesis
3 Methods
3.1 Latent Variate Factorial PCs
3.2 Principal Components
3.3 Machine Learning Unsupervised PCA and Clustering Algorithmic Techniques
4 Results and Discussion
5 Conclusion
References
Color Image Retrieval with a Weighted Adjacent Structure Model
1 Introduction
2 Related Work
3 Results and Discussion
4 Conclusion
References
Spam and Ham Classification by Multinomial Naïve Bayes Classification in Text Data
1 Introduction
2 Related Work
3 Implementation of the Work
4 Results
5 Conclusion
References
Author Index

Citation preview

Algorithms for Intelligent Systems Series Editors: Jagdish Chand Bansal · Kusum Deep · Atulya K. Nagar

Vinit Kumar Gunjan P. N. Suganthan Jan Haase Amit Kumar   Editors

Cybernetics, Cognition and Machine Learning Applications Proceedings of ICCCMLA 2021

Algorithms for Intelligent Systems Series Editors Jagdish Chand Bansal, Department of Mathematics, South Asian University, New Delhi, Delhi, India Kusum Deep, Department of Mathematics, Indian Institute of Technology Roorkee, Roorkee, Uttarakhand, India Atulya K. Nagar, School of Mathematics, Computer Science and Engineering, Liverpool Hope University, Liverpool, UK

This book series publishes research on the analysis and development of algorithms for intelligent systems with their applications to various real world problems. It covers research related to autonomous agents, multi-agent systems, behavioral modeling, reinforcement learning, game theory, mechanism design, machine learning, meta-heuristic search, optimization, planning and scheduling, artificial neural networks, evolutionary computation, swarm intelligence and other algorithms for intelligent systems. The book series includes recent advancements, modification and applications of the artificial neural networks, evolutionary computation, swarm intelligence, artificial immune systems, fuzzy system, autonomous and multi agent systems, machine learning and other intelligent systems related areas. The material will be beneficial for the graduate students, post-graduate students as well as the researchers who want a broader view of advances in algorithms for intelligent systems. The contents will also be useful to the researchers from other fields who have no knowledge of the power of intelligent systems, e.g. the researchers in the field of bioinformatics, biochemists, mechanical and chemical engineers, economists, musicians and medical practitioners. The series publishes monographs, edited volumes, advanced textbooks and selected proceedings. Indexed by zbMATH. All books published in the series are submitted for consideration in Web of Science.

Vinit Kumar Gunjan · P. N. Suganthan · Jan Haase · Amit Kumar Editors

Cybernetics, Cognition and Machine Learning Applications Proceedings of ICCCMLA 2021

Editors Vinit Kumar Gunjan Department of Computer Science and Engineering CMR Institute of Technology Hyderabad, Telangana, India

P. N. Suganthan Department of Electrical and Electronics Engineering Nanyang Technological University Singapore, Singapore

Jan Haase Nordakademie Elmshorn, Germany

Amit Kumar BioAxis DNA Research Centre Hyderabad, Telangana, India

ISSN 2524-7565 ISSN 2524-7573 (electronic) Algorithms for Intelligent Systems ISBN 978-981-19-1483-6 ISBN 978-981-19-1484-3 (eBook) https://doi.org/10.1007/978-981-19-1484-3 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

Preface

This book consists of selected papers from 3rd edition of International Conference on Cybernetics, Cognition and Machine Learning Applications organized on August 21–22, 2021, in India. Focused contents on smart machines, intelligent computing, communication technologies are selected in this edition based on the increased usage of such technologies in Industries for offering services and developing products. Data has become paramount for accuracy in innovations and engineering solutions these days, and some papers discussing the utilities in sectors like banking, insurance and other commercial applications were part of this year’s conference which generated interesting know-hows. Couple of chapters related to sustainable engineering solutions have been placed discussing sensors and deep learning related areas. A comprehensive knowledge on recent advancements in cognitive science and machine learning techniques makes the book appropriate for researchers and scholars working in the area of cognitive and computational intelligence disciplines. Segmentation techniques related to present-day health care like diabetic retinopathy, cancers and COVID-19 make it beneficial for engineers working in the healthcare solutions area. Several literature on cryptography and other security techniques are included too. Present-day applications of IOT relating to disaster prediction, sentiment analysis, intelligence techniques based on emotions, road safety and improving the quality of life are placed to enlighten the research community working in the IOT area. A couple of chapter also discusses the usage of machine learning techniques for crop and yield production applications and how automating the traditional approach can actually benefit in the present time.

v

vi

Preface

All in all, the book is expected to be catering the interest of academicians working in the areas of intelligent engineering, cyberphysical system, communication engineering and computational intelligence-related disciplines. It will be equally beneficial for Industry professionals intending to improve their knowledge in the areas related to the theme of ICCCMLA 2021. Hyderabad, India Singapore Elmshorn, Germany Hyderabad, India

Vinit Kumar Gunjan P. N. Suganthan Jan Haase Amit Kumar

Contents

A Deep Learning Approach to Analyze Traffic Congestions for Effective Traffic Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . K. Sai Prasad and S. Pasupathy Building Energy Consumption Prediction Model Using Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Chavala Lalithya Rao, Kurapati Sainath Raju, Pragati Mishra, P. S. G. Aruna Sri, and V. A. Narayana An End-to-End GUI-Based Real-Time Attendance System by Training Annotated Facial Data on YOLO v3 Algorithm . . . . . . . . . . . M. V. D. Prasad, M. Lakshmi Anusha, Medha Swapnika Kidambi, K. D. S. R. S. H. Srivastav, and M. Teja Kiran Kumar A Multifactor Security Protocol for Wireless Payment-Secure Web Authentication Using Mobile Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Challa Venkata Pranith, Valiveti Lohya Sujith, Kolli Sai Kiran, Pulivarthi Goutham, and K. V. D. Kiran

1

9

19

29

IoT-Based Greenhouse Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . R. Agilesh Saravanan, Gowri Priya, Sai Nishanth, Praveen Sai, and Vasanth Kumar

39

Home Automation Using Telegram Bot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sree Vardhan Cheerla, V. V. N. Chakravarthy, K. KishoreBabu, and V. GopiRam

47

Application of IoT in Hospital Management . . . . . . . . . . . . . . . . . . . . . . . . . . Sree Vardhan Cheerla, Syed Inthiyaz, V. Subba Reddy, K. SaiSaketh, N. Kiran Bhavya, and M. Vinay Kumar

55

Vehicle Tracking by Using IoT Applications . . . . . . . . . . . . . . . . . . . . . . . . . . A. Revanth, R. Pavan Kumar, R. Jyosthna, M. Sridhar, and Surendra Kumar Bitra

65

vii

viii

Contents

Smart Vehicle and Smart Parking System Using IOT . . . . . . . . . . . . . . . . . S. Yoga Sasidhar Reddy, Ch. Amarnath, K. Surya Teja, M. Sridhar, and Surendra Kumar Bitra

71

Automatic Water Irrigation System Using IoT . . . . . . . . . . . . . . . . . . . . . . . P. Sanjan Miller, B. Sai Bhaskar Reddy, M. Govardhan Reddy, M. Sridhar, and Surendra Kumar Bitra

79

Development of Safety Monitoring for an IOT-Enabled Smart Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Harshitha, Ch. Manikanta Uma Srinivas, M. Eswar Sai, Krishnaveni Kommuri, and P. Gopi Krishna

87

Low-Cost ECG-Based Heart Monitoring System with Ubidots Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 A. Vamseekrishna, M. Siva Ganga Prasad, P. Gopi Krishna, P. Bhargavi, S. Rohit, and B. Tanmayi Design of Hybrid Soft Computing Techniques for Estimation of Suspended Sediment Yield in Krishna River, India . . . . . . . . . . . . . . . . . 113 Arvind Yadav, Sanjay Vishnoi, Pragati Mishra, Devendra Joshi, and Haripriya Mishra QR-Based Ticket Verification and Parking System . . . . . . . . . . . . . . . . . . . . 123 Mohammed Ali Hussain, K. Sree Varsha, K. Krishnamraju, K. Lavanya, and B. Chakradhar Design and Analysis of Uniformly Illuminated 8-Way Wilkinson Power Divider for L-Band Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 K. T. P. S. Kumar, Lakshman Pappula, Vankayalapati Sahiti, M. V. Sai Kalyan, Nallamalapu Pratapreddy, and Potharaju Vinay Kumar Comparative Analysis on Mulberry Leaf Disease Detection Using SVM and PNN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 Y. Rakesh Kumar, P. Satyanarayana Goud, and Sheelam Pravalika Thermal Analysis of Different Components on the PCB Using ANSYS Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 K. Sanjitha, V. Panduranga, and S. Mallesh Custom Handloom Mobile Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 K. Pushpa Rani, N. Sushma swaraj, K. Himabindu, S. Shashank, and B. Vara Prasad A Compact Orthomode Transducer in K-Band for Satellite Communications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 V. Santhosh Kumar, NageswaraRao Lavuri, and Abdul Subhani Shaik

Contents

ix

A Review on Stock Market Analysis Using Association Rule Mining . . . . 171 R. Venkateswara Reddy, K. Venkateswara Rao, M. Kameswara Rao, and B. P. Deepak Kumar Student Performance Assessment Using AI . . . . . . . . . . . . . . . . . . . . . . . . . . 185 K. L. S. Soujanya, Challa MadhaviLatha, M. Swathi, Ch. Mallikarjuna Rao, and Sridevi Sakhamuri Automatic Smoke Absorber and Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 Venu Adepu, V. Ranga Sai Kiriti, K. Veera Bhadra, N. Sai Deepak, and P. S. G. Aruna Sri Computational Intelligence in Subthalamic Nucleus Deep Brain Stimulation: Machine Learning Unsupervised PCA Tracking Method and Clustering Techniques for Parkinson’s Feature Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 Venkateshwarla Rama Raju, B. Anuradha, and B. Sreenivas Color Image Retrieval with a Weighted Adjacent Structure Model . . . . . 215 N. Koteswaramma and Y. Murali Mohan Babu Spam and Ham Classification by Multinomial Naïve Bayes Classification in Text Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 J. K. R. Sastry, P. Harika, Trisha Dubey, and Y. Vijay Ditya Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235

About the Editors

Vinit Kumar Gunjan is Associate Professor in Department of Computer Science & Engineering and Dean of Academic affairs at CMR Institute of Technology Hyderabad (Affiliated to Jawaharlal Nehru Technological University, Hyderabad). An active researcher; published research papers in IEEE, Elsevier & Springer Conferences, authored several books and edited volumes of Springer series, most of which are indexed in SCOPUS database. Awarded with the prestigious Early Career Research Award in the year 2016 by Science Engineering Research Board, Department of Science & Technology Government of India. Senior Member of IEEE, An active Volunteer of IEEE Hyderabad section; 2021 additional secretary; 2021 Vice Chairman—IEEE Computational Intelligence Society; volunteered in the capacity of Treasurer, Secretary & Chairman of IEEE Young Professionals Affinity Group & IEEE Computer Society. Was involved as organizer in many technical & non-technical workshops, seminars & conferences of IEEE & Springer. He was awarded with outstanding IEEE Young Professional award in 2017 by IEEE Hyderabad Section. P. N. Suganthan is Professor at Nanyang Technological University, Singapore, and Fellow of IEEE. He is Founding Co-Editor-in-Chief of Swarm and Evolutionary Computation (2010–), an SCI Indexed Elsevier Journal. His research interests include swarm and evolutionary algorithms, pattern recognition, forecasting, randomized neural networks, deep learning and applications of swarm, evolutionary, & machine learning algorithms. His publications have been well cited (Google scholar Citations: ~33k). His SCI indexed publications attracted over 1000 SCI citations in a calendar year since 2013. He was selected as one of the highly cited researchers by Thomson Reuters every year from 2015 to 2018 in computer science. He served as General Chair of the IEEE SSCI 2013. He is IEEE CIS Distinguished Lecturer (DLP) in 2018– 2020. He has been Member of the IEEE (S’91, M’92, SM’00, Fellow’15) since 1991 and elected AdCom Member of the IEEE Computational Intelligence Society (CIS) in 2014–2016.

xi

xii

About the Editors

Jan Haase (M’07, SM’09) received his Bachelor, Master, and Ph.D. degree in computer sciences at University of Frankfurt/Main, Germany. Then he was project leader of several research projects at University of Technology in Vienna, Austria, at the Institute of Computer Science and a lecturer at Helmut Schmidt University, Hamburg, where he received his habilitation grade. 2016–2020 he held a temporal professorship for Organic Computing at University of Luebeck, Germany and now is a full professor at Nordakademie near Hamburg, Germany. His main interests are Building Automation, System Specification and Modeling, Simulation, Low-Power Design Methodologies, Wireless Sensor Networks, Automatic Parallelization and modern Computer Architectures. As a member of several technical program committees of international conferences he is involved in the review process of many research publications and repeatedly acted as TPC chair, track chair, etc. at these conferences. He (co)-authored more than 100 peer reviewed journal and conference papers and several book chapters. As an IEEE volunteer, he currently is Germany Section’s Chair, has been Austria Section’s Chair and is active in IEEE R8, having been R8 Conference Coordinator, R8 Professional Activities Chair, and a member of several R8 committees. In IEEE Industrial Electronics Society, he is a voting AdCom member. On IEEE HQ level he currently serves on the Conference Finance Committee and chaired the Adhoc on Cultural Differences in the IEEE Conferences Committee. He also continuously served as a mentor on the IEEE VoLT program since the very first season. Amit Kumar Ph.D., A DNA Forensics Professional, Entrepreneur, Engineer, Bioinformatician and an IEEE Volunteer. In 2005 he founded the first Private DNA Testing Company Bio Axis DNA Research Centre (P) Ltd in Hyderabad, India with an US Collaborator. He has vast experience of training 1000+ Crime investigation officers and helped 750+ Criminal and non-criminal cases to reach justice by offering analytical services in his laboratory. His group also works extensively on Genetic Predisposition risk studies of cancers and has been helping many cancer patients from 2012 to fight and win the battle against cancer. Amit was member of IEEE Strategy Development and Environmental Assessment committee (SDEA) of IEEE MGA. He is senior member of IEEE and has been a very active IEEE Volunteer at Section, Council, Region, Technical Societies of Computational Intelligence and Engineering in Medicine and Biology and at IEEE MGA levels in several capacities. He has driven a number of IEEE Conferences, Conference leadership programs, Entrepreneurship development workshops, Innovation and Internship related events. Amit was Chairman of IEEE Hyderabad section in 2020 and He was visiting Professor at SJB Research Foundation till 2020. He was awarded the prestigious IEEE MGA Achievement award in 2020 for the outstanding contributions at Section, Council and Region levels. Currently He is member of IEEE MGA Nominations and Appointments committee and Chairman of IEEE Region 10 Section Chapter committee. He was winner of IEEE MGA Achievement award and IEEE India council outstanding volunteer award in 2020.

A Deep Learning Approach to Analyze Traffic Congestions for Effective Traffic Management K. Sai Prasad and S. Pasupathy

1 Introduction As the population of humans are increasing day by day, parallelly, the vehicle’s production companies and vehicles production rate are also getting increased, and with these roads, expressways and highways are becoming overcrowded [1]. To make transportation safer, reliable, cleaner, and efficient, Intelligent Transportation System (ITS) is applied to assemble, cognize, and manage the information regarding the transportation flows from bases, and these days, it is also getting important to keep track of information about the vehicles on the roads, highways, and expressways [2]. To keep track of vehicles, there are some methods like inductive loop detector which can be traditionally achieved, an infrared detector, radar detector, and videobased solution. Among those techniques, the video-based solution is the best choice because it does not cause any disturbance for the traffic flow, and also, it will not face any problems in weather conditions, whereas if we go to radar or infrared detectors, those are based sensors, and these cannot detect the vehicles accurately when it is raining and when a vehicle shadow falls on other vehicles, whereas in the video-based solution, we can avoid these type of difficulties and also we can get the videos from a surveillance camera mounted outdoor, and in these days, on every road, cameras are mounted and getting captured regularly, and those videos can be used for this project, so there is no need of installing anything from scratch [3].

K. S. Prasad (B) Department of Computer Science and Engineering, MLR Institute of Technology, Hyderabad, India e-mail: [email protected] S. Pasupathy Department of Computer Science and Engineering, Annamalai University, Chidambaram, Tamil Nadu, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_1

1

2

K. S. Prasad and S. Pasupathy

As there are many advantages of video-based solution like traffic flow was undisturbed and no need of installing anything from scratch, easily installed and also it can be modified whenever there is a requirement and etc.., because of all these advantages, it has drawn the noteworthy courtesy in the past decade by the researchers [4]. On road, the vehicles will be moving constantly, so for that, background subtractor should work dynamically, and for that, we can use OpenCV background subtractor which detects the moving vehicles as the objects removes the noise and shadows dynamically; by using this, we can eliminate the noise and shadows so that the vehicles behind the heavy vehicles and the vehicles under the shadow of other vehicles can also be detected and make our result more accurate [5]. To track and detect the moving vehicles is a crucial task in this project. Here, we define two techniques in which the problem is condemned quite differently. The one among it is the virtual detector, and the other method is blob tracking. Using any one of these methods, we can detect the moving vehicles. The method which we used in this project is the virtual detector, and what it does is from the surveillance video; a rectangular frame is pointed to the vehicles, and those rectangular frames will be treated as vehicles.

2 Existing Features At the beginning to keep track of vehicles, they use to count the vehicles going on the road manually which is a hectic task to count the vehicles manually, and the accurateness of the result is also very low, and as the technology is increasing, they used sensors to keep track of vehicles going on roads [6]. Some of those sensorbased techniques are infrared detector, radar detector. In this, the major drawbacks are: It cannot detect the vehicles accurately when the vehicles are covered by heavy vehicles and when it is raining, etc. [4]. In the present generation, video-based solutions are used which provide more accurateness than previously used techniques, and some of the techniques used in the video-based solution are • • • •

Point detection. Edge detection. Frame detection. Matching.

3 Proposed System The technique used in this paper is different from other techniques, and to provide better quality and accuracy in vehicle detection, it uses a hybrid technique; that, is it is a combination of both “frame differentiation” and “edge detection” algorithms,

A Deep Learning Approach to Analyze Traffic …

3

and we are using the Kalman filter with which the position of each vehicle can be assessed and tracked properly and accurately and also provide useful information for traffic flow analysis; this Kalman filter is used to organize the different vehicles in dissimilar definite clusters and count them distinctly. Motion analysis using a combination of different techniques, edge detection, detection zone definition, image enhancement process, and vehicle-type sorting and counting, all these methods are included in this technique, and it is essential to say that some norms made in this work like there is no sudden change in direction of vehicles are expected, and motion scenes which are given to our project should be captured with a view from the overhead of roadway surface. Advantages • Noises are eliminated from the surveillance video photograph. • Shadows will be removed from the input motion so that even the vehicles under the shadow of other vehicles can also be detected and tracked. • We can use the output of the project for traffic management like if in a junction, there are four sides to vehicles pass by, and the number of vehicles pass by two sides of them is very less; then, using some IoT tools, we can statically operate the signals at those junctions and make the traffic flow efficient.

4 Methodology Step 1: Our program will take the input video file from the users and capture the given input video file and the properties of that corresponding video file. Step 2: An imaginary line is generated to our video file. Step 3: With the help of an OpenCV object, a background subtractor is applied to our video file, and the noise and shadows will be eliminated from our input video file. Step 4: The Kalman filter is applied with which the position of each vehicle can be assessed and tracked properly and accurately. Step 5: The moving vehicle will be treated as an object, and a frame will be initialized to it and a point to the center of that frame.

4

K. S. Prasad and S. Pasupathy

Fig. 1 Block diagram

Step 6: Whenever a vehicle with frame and point crosses the imaginary line, the count of number of vehicles passed by will be incremented (Figs. 1 and 2).

5 Results See Figs. 3, 4, 5, 6, 7, and 8.

6 Conclusion and Future Work The proposed system is implemented using Python OpenCV. In this, the video footage which is recorded from the overhead of the roadway surface that is from the traffic surveillance video is given as input, and our project will apply the image processing technique, background subtractor, and the Kalman filter to calculate the total number of vehicles passed by that road and gives us the number of vehicles passes by. Currently, our proposed project is processing on already captured videos, but it can be modified to live processing of video; that is, by adding microcontrollers, we can attach our technique to video capturing cameras and can store the data we get directly to databases and can be retrieved whenever we need it, and also we can manage the traffic statically and efficiently, and at the toll gates based on our output, they can open and close the tollgate booths based on the vehicles pass by. One of the limitations of our project is it does not detect the stopping up of the vehicles which affects the accuracy of our result. Another limitation of the current system is we define an imaginary line, and when the vehicles’ frames points intersect our imaginary line, we consider it as the vehicle passed by; if that imaginary line is placed at the wrong place in our video, then it

A Deep Learning Approach to Analyze Traffic … Fig. 2 Flowchart

Fig. 3 Result of background subtractor

5

6 Fig. 4 Gray color region shows shadow region

Fig. 5 Result of masked image

Fig. 6 Result of masked image2

K. S. Prasad and S. Pasupathy

A Deep Learning Approach to Analyze Traffic …

7

Fig. 7 Capturing the vehicles

Fig. 8 Final output

leads to wrong output, so to overcome this problem, an imaginary line placer which is developed based on artificial intelligence can be used in the next generation of this project.

References 1. Fathy M, Siyal MY (1995) An image detection technique, based on morphological edge detection and background differencing for real-time traffic analysis. Pattern Recogn Lett 16:1321–1330

8

K. S. Prasad and S. Pasupathy

2. Sethi K, Jaiswal V, Ansari MD (2020) Machine learning based support system for students to select stream (subject). Recent Adv Comp Sci Commun (Formerly: Recent Patents on Computer Science), 13(3):336–344 3. Prasad PS, Pathak R, Gunjan VK, Ramana Rao HV (2020) Deep learning based representation for face recognition. In: ICCCE 2019. Springer, Singapore, pp 419–424 4. Syed AT, Merugu S, Kumar V (2020) Augmented reality on sudoku puzzle using computer vision and deep learning. In: Advances in cybernetics, cognition, and machine, Lecture notes in electrical engineering. Springer, Singapore, pp 567–578 5. Prasad KS, Miryala R (2019) Histopathological image classification using deep learning techniques. Int J Emerg Technol 10(2):467–473 6. Prasad KS, Pasupathy S Dr. Deep learning concepts and libraries used in image analysis and classification. In: TEST engineering and management, vol 82, pp 7907–7913. ISSN: 0193-4120

Building Energy Consumption Prediction Model Using Machine Learning Chavala Lalithya Rao, Kurapati Sainath Raju, Pragati Mishra, P. S. G. Aruna Sri, and V. A. Narayana

1 Introduction Solar energy is the purest and most plentiful renewable energy source handy. Solar endures within a compact and interrelated ignition system, operating adjacent to other technologies to a reliable energy austerity. Solar energy is a very resilient energy technology, and it can be produced as dispersed generation or as a fundamental station, convenience-scale solar power plant (like traditional power plants) [1]. Using cutting-edge solar and storage systems, some strategies may store the energy they generate for sharing during sunsets [2]. Solar technologies can provide energy for various applications, including producing electricity, rendering light or a comfortable in land environment, and heating water for residential, industrial, or everyday use. Solar energy applications depend on supportive system cores at the provincial, federal, and national levels to assure customers, and industries have decent access to reliable energy technologies like solar [1]. Everyone can relish solar energy in several distinctive ways: • Photovoltaic cells: These cells can transform sunlight into electricity. • Solar thermal technology: The sunlight can provide us with the means to make hot water or steam using solar thermal technology. • Passive solar heating: It can be as easy as subletting the sunshine by windows to ignite the building’s core. This essay includes a taxonomy overview of emerging artificial intelligence-based solar power forecasting models. Based on their variations and similarities, taxonomy distinguishes solar energy prediction approaches, optimizes, and forecasts systems C. L. Rao · K. S. Raju · P. Mishra · P. S. G. A. Sri (B) Department of ECM, KLEF, Vaddeswaram, Andhra Pradesh 522503, India e-mail: [email protected] V. A. Narayana Department of CSE, CMR College of Engineering and Technology, Kandlakoya, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_2

9

10

C. L. Rao et al.

into many groups. It also demonstrates the difficulty and underlying characteristics of study inclinations in solar power forecasting using AI algorithms [3, 4]. It theoretically examines the aspects of several solar prediction models, whereby assisting us in picking the most suitable model in the application outline. Therefore, as energy is sustainable and renewable for the future, our principal focus is to study the various energy consumption prediction models available. Concerning those, we will be predicting a prediction to strengthen the accuracy levels [4]. The main goal of the study will be based upon the prediction of solar forecasting using artificial intelligence, machine learning so that we can expect the power production in various weather conditions. As a result, improving solar power prediction accuracy is critical in order to plan for unanticipated potential events [3, 5]. Since they can reveal the invariant composition and nonlinear features in solar results, artificial intelligence algorithms have been widely publicized with contentious prediction enforcement.

2 Literature Review Creating an energy use forecast is an important goal of energy planning, management, and conservation [6]. The data-driven model is probably the most cost-effective approach for forecasting building energy usage. The aim of this research was to find an energy consumption prediction model with high precision, accuracy, and generalization capability. The conclusions indicate that the suggested approach produces better performance than generally used systems concerning efficiency, accuracy, and robustness, compared to state-of-the-art standards and the regularly used datadriven prototypes. The recommended method assures high efficiency, a generalized technique, and accuracy for building energy consumption predictions. Machine learning is currently one of the most common methods for hourly star prediction [7]. Efficiency is dependent on variables such as temperature and atmosphere, and there is no uniform model. In this analysis, a total of 68 machine learning models was tested using satellite-derived irradiance data from seven locations around five separate environmental zones on the mainland of the United States. But for RFqr under overcast skies, none of the approaches is dominant in terms of numbers. The research discovered that tree-based methods outperformed all-sky methods in terms of long-term average name. PV power [1] engendering forecasts must be accurate in order to incorporate the fickle energy source into the power grid. New models can be tested with data from different environmental areas, as well as proper benchmarking. Building performance is progressively sensitive to tenant behaviors. Informationdriven strategies are used to predicting the building’s energy utilization with the restricted physical data [6]. It is well recognized that the forecast of energy utilization in the large, medium, and short term is essential for energy market planning and investments. A CO2 -detection sensor network was constructed to calculate the

Building Energy Consumption Prediction …

11

gas concentration in the building. The concealed parameters, including the occupancy and design parameters, are determined for using a frequent maximum likelihood algorithm and Bayesian estimation. Predicting the energy utilization of the air conditioner is an integral part of the forecasting the total building energy utilization [8]. The study can help develop a forecast controller of HVAC systems for energy management and thermal comfort [2, 6]. Service organizations could apply request side management that calculates after taking out the load for forecasting to reach a proper load-shape objective, such as load shifting. This forecasting process for building energy prediction is applied to original cases as a demonstration, and the results verify its accuracy [9]. The building and construction area are responsible for 36% of the global final energy consumption and nearly 40% of the total direct and indirect carbon dioxide emissions [10]. Building energy storing can be attained by improving the building’s dynamic energy execution in the feasible development management in the urban environment building. Anwell organized modeling technique which is developed for solar power prediction, and the MKRVFLN technique proved to forecast solar power more precisely than some traditional methods. The intended EVWCA-MKRVFLN method provides better reliable predicting results even with the intervention of outliers. In the forthcoming trial, several atmospheric limitations can be considered for validating the performance of the intended technique [11]. Building thermal load forecasting informs the enhancement of cooling plants and thermal energy storage [12, 13]. Building load forecasting has different applications in the fields of HVAC control, thermal storage operation, smart grid management, and others [11, 14].

3 Background 3.1 Artificial Intelligence The intensity of a digital computer or computer-controlled robot to perform duties is typically incorporated with intelligent individuals. The concept is likewise used in the project of intending edifices endowed with humanistic intellectual mechanisms, such as the ability to reflect, examine the application, theorize, or avail from former experience. Since, the design of the digital computer has been explicated that machines can be programmed with significant and considerable dexterity to accomplish very intricate responsibilities, such as attaining data for scientific theories. Notwithstanding constant advancements in computer processing velocity and memory power, there are yet no systems that can match human versatility across higher realms or fields that require a lot of quotidian data. On the other hand, in administering such critical purposes, several algorithms have outdone the product standards of human specialists and scholars, such that AI in this linear setting can be used in applications as

12

C. L. Rao et al.

diversified as therapeutic care, computer quest engines, and identification of speech or handwriting [15].

3.2 Machine Learning Machine learning is an AI technology that concedes programs to study and progress from experience automatically without being programmed undeviatingly. The focus of ML is on outlining computer programs that can operate and use data and study regarding themselves [16]. The learning process commences with the perspicacity to hunt for correspondences in data and obtain versed selections in the future based on the standards we have, such as examples, linear knowledge, or guidance. The fundamental purpose is to concede computers to learn and adapt behavior without human involvement or embroilment automatically. ML algorithms are also categorized as supervised and unsupervised by specific ML techniques. When applied to AI, there are a diversity of multiple learning techniques. The machine should then maintain the solution with the situation to identify the solution the next time the exemplar discerned the corresponding position. It is reasonable to fortify this primary memorization of individual articles and processes, known as rote learning, on a computer [17]. The issue of utilization of what is estimated generalization is likewise challenging. Generalization signifies acclimating the former experience to alike current outlines. A program that learns the past tense of conventional English verbs by rote, for instance, would not be able to determine the past tense of a term such as ‘jump’ till it was beforehand was rendered with ‘jumped’. In contrast, a program that will generalize can perceive the ‘added’ rule and thus compose the past tense of ‘jump’ based on similar verbs conversance [18].

3.3 Linear Regression It is a linear model, for example, a model that considers a linear correlation among the input variables (i.e., x) furthermore individual output variables (i.e., y). More concretely, the output variables can be ascertained from a linear succession of the input variables (which are x and y).When we have an exclusive input variable (i.e., x), the process is indicated being a simple linear regression. Meanwhile, there is various distinct input variables, and analysis from statistics often relates to the multiple linear regression method. When there is a single input variable (x), the process is indicated as a simple linear regression. When there are various input variables, analysis from statistics often relates to the multiple linear regression method. Several techniques can be utilized to develop or train the linear regression equation from data, and the most prevalent is termed as ordinary least squares. It is customary to consequently point to a model developed this way as least squares regression [5]. Learning a linear

Building Energy Consumption Prediction …

13

regression model involves predicting the values of the coefficients utilized in the description with the data that we have ready. Linear regression has been analyzed to a vast extent, and there is a lot of discussion on how the data must be structured to gain the most beneficial application of the model.

4 Methodology In this power consumption system, the energy is obtained from the solar energy system because it is renewable energy, and it is a pollution-free energy generation system. In this project, we use different AI algorithms to prognosticate energy. This study is about solar power forecasting by using various prediction methods with the help of artificial intelligence [4]. So, we can estimate the power generation in infrequent weather conditions. The primary distinction of this research is we can easily detect solar radiation with the help of a detector (pyranometer). Still, we oblige to prophesy the power with the aid of artificial intelligence [5]. Thus, it would be obvious to maintain and estimate. We ordinarily applied one of the mere AI algorithms, linear regression, for prophesying the temperature. Regression analysis is a form of prediction modeling technique that procures the correlation between dependent and independent variables. Here, we have taken temperature as dependent and time as an independent. The objective is forecasting, predicting, or fallacy minimization; linear regression would be used to harmonize a predictive model to a scrutinized dataset of values of the response and explanatory variables. After developing the sort of model, if the explanatory variables’ supplementary values are accumulated without an accompanying response value, the outfitted version can be used to predict the response. With linear regression, we can minimize the error percentage between the original value and the predicted value. Therefore, with help of a predicted value, we can calculate how much energy is generated. The mathematical relation between the temperature and time is as follows:

Impressions for a way ANN functions: • It assists in conjecture the impression of progressing/contracting the dataset vertically or horizontally on computational time. • It aids to surmise the states or circumstances, where the model fits the ablest. • It also supports to articulate why a definite model operates more suitable in particular environments or circumstances. ANN is seldom utilized for predictive modeling. The idea prevailing that ANN customarily adjudicates to overfit the association. ANN is typically used in states where past events have a strong tendency to repeat themselves. For example, suppose we are playing the match of Black Jack opposite to a machine. An intellectual

14

C. L. Rao et al.

Fig. 1 Error percentage for the above linear regression graph is 0.8, and it is most accurate. The above graph is the relation between temperature, which is denoted with the dependent variable, and the time, denoted with an independent variable

contender based on ANN would be a pretty great competitor in this position, considering they can contrive to grip the reckoning time flat. ANN will train itself for all probable quandaries of card flow, and provided that we are not shuffling cards with a merchant, ANN will be prepared to record every single note. Therefore, it is a sort of ML technique that has tremendous memory. But, it does not operate well in situations where the scoring community is significantly altered compared to the training sample. For example, if we strategize to aim clients for a crusade using their preceding response by an ANN. We will presumably be adopting an obverse method as its sways have overfitted the correlation between the rejoinder and others (Figs. 1, 2, 3 and 4).

5 Results A broad and orderly survey of fabricated consciousness-based solar-oriented force expectation is unexpectedly led from the perspective of taxonomy. A great deal of measurable investigation of flow research interests and examination hotspots

Building Energy Consumption Prediction …

15

Fig. 2 Training: The sample of the data is used to fit the respective model. A set of patterns is used for the learning, which implies to provide parameters (weights) of that classifier. The training is completed between the ambient temperatures

Fig. 3 Validation: Sample data applied to render an impartial evaluation of a model matches the training dataset also while harmonizing model hyperparameters. The evaluation matures more biased as a facility on the validation dataset is consolidated into the model contour

16

C. L. Rao et al.

Fig. 4 Testing: Sample data that is applied to render an impartial evaluation of a terminal model that fits the training dataset. And hereabouts, we can also find the error percentage between the original data and predicted data. To find the error percentage for the linear regression graph, we used the R-square method

for solar-oriented force forecast is completed. The scientific categorization gives a request characterization of manufactured consciousness, the technique, streamlining agent, and forecast structure as indicated by their standard qualities and connections. A few difficulties and future exploration bearings in sunlight-based force forecast dependent on synthetic consciousness are additionally given. Our examination shows that every AI strategy, enhancer, and forecast structure has its preferences and disservices. These examinations and mathematical introductions can help sun-based energy experts, for example, researchers and specialists, to figure out which human-made consciousness calculations and forecast structures can improve their particular expectation apparatuses, consequently assisting with exploring the capability of man-made consciousness in solar-based force determining. We outline our work on solar power forecasting by using various prediction methods with the sustenance of artificial intelligence algorithms. So, we can estimate the power generation in various weather conditions. The primary difference in our work is that we can readily detect solar radiation with the help of a detector (pyranometer); however, we oblige to forecast the power with the lieutenant of artificial intelligence. The error percentage we perceived for the linear regression graph is 0.8.

Building Energy Consumption Prediction …

17

Furthermore, we are also improving the error performance and efficiency to acquire more enhanced accuracy on the result.

6 Conclusion Machine learning methods have, in the recent past, made a very important contribution to the progression of the forecast models adopted for energy dissipation and consumption. Such models profoundly enhance the efficiency, accuracy, robustness, precision, and generalization capability of the conventional timeline forecasting mechanisms. Our study analyzes the state-of-the-art models adopted in the customary utilization of energy consumption. The most suitable, by discovery and taxonomy, in the domain, is graded according to the modeling technique, energy model, condemnation form, and the realm of the application. A comprehensive analysis of the relevant machine learning techniques is classified in literature, their usage, and a study on assessing their effectiveness in predicting solar energy consumption. This article makes an upshot of the tendency and the performance of the models. In conclusion, the analysis relates to a major development in the accuracy and ever-increasing performance of the forecast using the prediction models. A comprehensive and systematic study of AI-based solar power forecast holds plenty of analytical reviews of prevailing study concerns and hotspots for solar power predicting. Numerous hurdles and future analysis incline toward solar power predicting based on ML. This study illustrates that each prediction model has its benefits and drawbacks. These studies and statistical portrayals can help solar energy practitioners determine which algorithms and predicting models can enhance the explicit predicting tools, whereby examining ML in solar power predicting. Also having several analyses on the papers about solar power predicting, however, still there are a few index concerns that have nonetheless been conclusively resolved to be very intricate. Furthermore, not only associated with the practical working factors of the PV cells still further associated with the environmental determinants. Hence, the efficient predicting model is 1 of the important accurate enigmas been determined in the coming days. Currently, nearly full records usually express a dilemma of solar energy predicting. In brief, the solar power predicting model is 1 of the individual principal study inclinations in the coming days. Also, several subproblems in PV power predicting, such as the respective predictions: point, multi-step, day-ahead, and probability. Meanwhile the prevailing discussion, certain subproblems are not only economically and individually solved. Various prediction responsibilities are a significant quandary that the scholarly association requires to solve necessarily.

18

C. L. Rao et al.

References 1. Tang S et al (2018) Efficient path of distributed solar energy system synergetically combining photovoltaics with solar-syngas fuel cell. Energy Convers Manage 173:704–714 2. Wang Z, Hong T, Piette MA (2020) Building thermal load prediction through shallow machine learning and deep learning. Appl Energy 263:114683 3. Wang H et al (2020) Taxonomy research of artificial intelligence for deterministic solar power forecasting. Energy Convers Manage 214:112909 4. Shaik AS, Karsh RK, Islam M (2021) Robust image hashing using chromatic channel. In: Nath V, Mandal JK (eds) Proceeding of fifth international conference on microelectronics, computing and communication systems. Lecture notes in electrical engineering, vol 748. Springer, Singapore 5. Kumar KR, Kalavathi MS (2018) Artificial intelligence based forecast models for predicting solar power generation. Mater Today: Proc 5(1):796–802 6. Nam S, Hur J (2019) A hybrid spatio-temporal forecasting of solar generating resources for grid integration. Energy 177:503–510 7. Zhong H et al (2019) Vector field-based support vector regression for building energy consumption prediction. Appl Energy 242:403–414 8. Yagli G, Yang D, Srinivasan D (2019) Automatic hourly solar forecasting using machine learning models. Renew Sustain Energy Rev 105:487–498 9. Sethi K, Jaiswal V, Ansari MD (2020) Machine learning based support system for students to select stream (subject). Recent Adv Comput Sci Commun (Formerly: Recent Patents on Computer Science) 13(3):336–344 10. Merugu S, Reddy MCS, Goyal E, Piplani L (2019) Text message classification using supervised machine learning algorithms. In: Kumar A, Mozar S (eds) ICCCE 2018. Lecture notes in electrical engineering, ISSN 1876-1100, vol 500. Springer, Singapore 11. Zhao Y et al (2016) A novel bidirectional mechanism based on time series model for wind power forecasting. Appl Energy 177:793–803 12. Wang R, Lu S, Feng W (2020) A novel improved model for building energy consumption prediction based on model integration. Appl Energy 262:114561 13. Majumder I, Dash PK, Bisoi R (2019) Short-term solar power prediction using multi-kernelbased random vector functional link with water cycle algorithm-based parameter optimization. Neural Comput Appl:1–19 14. Rashid E, Ansari MD, Gunjan VK, Khan M (2020) Enhancement in teaching quality methodology by predicting attendance using machine learning technique. In: Modern approaches in machine learning and cognitive science: a walkthrough. Springer, Cham, pp 227–235 15. Fan C, Xiao F, Wang S (2014) Development of prediction models for next-day building energy consumption and peak power demand using data mining techniques. Appl Energy 127:1–10 16. Bourhnane S et al (2020) Machine learning for energy consumption prediction and scheduling in smart buildings. SN Appl Sci 2(2):297 17. Mosavi A, Bahmani A (2019) Energy consumption prediction using machine learning: a review 18. Prasad PS, Pathak R, Gunjan VK, Ramana Rao HV (2020) Deep learning based representation for face recognition. In: ICCCE 2019. Springer, Singapore, pp 419–424

An End-to-End GUI-Based Real-Time Attendance System by Training Annotated Facial Data on YOLO v3 Algorithm M. V. D. Prasad, M. Lakshmi Anusha, Medha Swapnika Kidambi, K. D. S. R. S. H. Srivastav, and M. Teja Kiran Kumar

1 Introduction Facial recognition is the best new technologies in the world. The technology that is most effective is recognizing an individual. Humans have different ways to identify a person; one of the technologies is biometrics. As technology is developing with the present era to make it more convenient and without any alleviation of time, facial recognition is introduced. A person can be identified very easily without any cooperation of that person like a posture, wearing an ID card, etc. In this documentation, the algorithms that we use in machine learning and deep learning play a crucial role. Generally, facial recognition is useful in many other ways like unlocking phones, which is one of the finest features in a phone. This feature is available in iPhones and some of the android phones like Samsung, which helps us to save personal data even when our phone is stolen. And at recent times, Listener has developed an app that helps blind people to recognize the other person’s feelings. This app recognizes when a person is smiling and then buzzers the vibrating blind person, that can offer the blind an awareness of social situations. Facial recognition is useful in universities, but all the corporate universities are taking attendance manually like recording the attendance of the student in a book or saving the data of a student in a laptop. It alleviates the precious time for the faculty as well as the students, and so person recognition can be used. It is more relaxed and convenient for students and the faculty. Machine learning is the analysis of algorithms that progressively evolve over experience. It is considered as a subclass of artificial intelligence. To make snap decisions without being explicitly trained to do so, machine learning algorithms M. V. D. Prasad (B) · M. L. Anusha · M. S. Kidambi · K. D. S. R. S. H. Srivastav · M. T. K. Kumar Department of ECE, KLEF Deemed to be University, Vaddeswaram, Andhra Pradesh 522502, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_3

19

20

M. V. D. Prasad et al.

construct a statistical equation based on sample data, known as “training data.” The border family of machine learning comes from deep learning which is based on ANN. Certain algorithms like YOLO, CNN, and ANN which are based on ML and DL are often used to boost detection performance, productivity in identifying the right individual, and recognition of specific individuals. YOLO algorithm is one of the most efficient algorithms, it is a real-time object detection algorithm, and we can know where and what the objects are in the image. This algorithm corresponds to a full image of a single neural network and then breaks the image into different regions or segments, estimating bounding boxes and possibilities for each and every region or segment. Basically, the aim of this experimental process is to make use of a productive method in identifying the students in a university, take advantage of their involvement from the students on a regular basis, and retaining the daily report. Face recognition applications could then be used to identify the student’s presence, which will solve the issues we are facing like taking attendance manually and to avoid proxies in the universities and schools. The model “Machine learning based 2D pose estimation model for human action recognition using geometrical maps” stipulated by M. V. D. Prasad, K. Durga Bhavani, S. M. Tharun Kumar, and P. V. V. Kishore states that due to its many uses such as security cameras, human machine interaction, and video recovery, identification of human interaction has become a more evolving subject in computer vision. Later, as opposed to computer vision algorithms, human action recognition obtained greater performance using machine learning algorithms. The methodology of action recognition was improved in the next phase, and further emphasis was placed on its potential to segment a human body and pinpoint joints. In this stipulated model, they are using pose estimation to derive geometric characteristics, to establish a human motion recognition system, and these characteristics are inputted into a sequential convolution neural network to recognize the subjects’ behavior.

2 Literature Review A model stipulated by Gandhe [1] defines the face recognition way to identify the person using different techniques. This system gives conformation to the system by face as a key to access. This system gives information about different things like identification of the person and access control. It says the face recognition is an easier way to access. Aditya Ladage [2] had introduced a new way of attendance monitoring by the use of mobile phones available with the teachers and students. YOLO algorithm was used for face detection. Their system will automatically mark student’s attendance that saves teachers time. Naveed et al. [3], The two databases contain the images of the students, along with their corresponding registration numbers, and the another is used to capture the attendance of the students present in the class. Firstly, the camera will take the

An End-to-End GUI-Based Real-Time …

21

pictures of the student with a clear background and without any disturbances, and then, it will compare with the student database and the attendance will be saved in the other database. Riddhi Patel [4] presumes that for the data and information there is much need for high security. Face biometrics is used for a person’s identification which is a simple way that shows the face in all dimensions and makes the 3d model on the computer. They presented a view of face recognition and applications. Nirmalya Kar [5] postulates that authentication is the main issue in the control of computer-based communication. Person facial recognition is the subject of user authentication and is employed in a variety of settings, including video monitoring, door control, and information security. This paper offers an overview of the approach used by the student’s attendance system that will use the YOLO algorithm to use face recognition technology. This device will automatically record the student’s attendance in the classroom. Manikandan [6], Research into the creation of facial recognition technologies for diverse uses was spearheaded by the introduction of elevated sensors, as well as elevated processors. Based on the application, asynchronous information or actual input is used in face detection technologies. The architecture and assessment of a real-time face recognition system using the CNN are suggested in this article. Using regular AT&T datasets, the preliminary assessment of the structural architecture is completed and the same is later applied to the real-time device configuration. Information is also recorded on the adjustment of CNN parameters to test and boost the accuracy of the proposed system’s identification. To boost the system’s performance, a systematic approach to tuning the parameters is also suggested. Total identification accuracies of 98.75% and 98.00% are achieved for regular datasets and real-time inputs, respectively, by using the proposed method. Sucharith [7], The role of the proposed system in the report, facial recognitionbased attendance system [5], was to collect the facial images of each and every student and hold it for their attendance in the repository. The camera is positioned inside a classroom at a particular distance to capture footage of the frontal pictures of all the students in the class. Frames from the video should be separated: For better identification and analysis of the students encountered, the recorded footage must be transformed into frames per second to create the attendance monitoring database network: CNN. It may be introduced in wider environments, such as in a lecture space, where it allows us to feel multiple people’s involvement. Often, the poor lighting state of the classroom can indirectly undermine the efficiency of the device, which can be solved by using certain algorithms or the quality of the film. Ade Kurniawan [8, 9], In studying, student participation is necessary. Oh, method. Several approaches can be used to monitor student attendance. Completed; one of them is by signatures from pupils. There are some weaknesses, such as taking a long time to create participation; the attendance log is missing; and the administration attendance data must be inserted into the machine one by one. To resolve this, the article recommended a web-based student system for attendance that uses facial recognition. The recommended one CNN method is used for detecting deep metric learning, faces in pictures, is used to generate facial embeddings and K-NN is used

22

M. V. D. Prasad et al.

to identify the faces of pupils. Accordingly, your machine can identify images. From the exams when carrying out, the machine was able to recognize the faces of students who attended, and their details on attendance are saved immediately. The university leadership, thus, is recording attendance data which has been limited [10, 11].

3 Methodology YOLOV3 algorithm—YOLO (You Only Look Once) is an algorithm used to detect artifacts in real time. As compared to the R-CNN group, YOLO is one of the fastest algorithms. The basic idea of YOLO is it consists of placing a grid on the image. YOLO algorithm is mainly classified into 2 groups. The first phase is divided into 2 stages, and the first phase is used for selecting important regions from the picture, while we identify those regions with the help of CNN in the second phase. The typical examples for the first category algorithms are area-based convolutional neural networks and fast R-CNN. Second group of algorithm concentrates on the regression. In such algorithms, instead of selecting interesting components, we attempt to anticipate categories and video frames for the full picture in the image in one action [12, 13]. R-CNN is very precise, focusing on parts of the image that might be more likely to contain an object rather than looking at the whole image. R-CNN’s main issue is that it is very sluggish and that it is not a full end-to-end object detector. Second stage detectors are more accurate when compared to single-stage detectors. The YOLO algorithm can process up to 45FPS on a GPU, while only 5FPS on a GPU can be processed by R-CNN. YOLO takes a picture of the input. The picture is sub-divided by the YOLO system into square grids. On each of the square grids, image classification and localization are then applied. Finally, the bounding boxes and their class probabilities are predicted by YOLO to detect the object if an object is identified. Currently, the YOLO implementations are done into three: dark net, AlexyAB/dark net, and the other one is dark flow; each one has their own advantages and disadvantages. Among all these, dark net was successfully implemented by the people behind this algorithm [14, 15]. In this article, we will be using an openCV and Python, and in this, the object will be detected in a still picture the applying one of the accurate algorithms called YOLO. We will be using an openCV which is a computer vision library or framework that will support our YOLOV3 algorithm. This openCV has inbuilt support for the dark net. We use dark net architecture because it is a pre-trained model which classifies 80 different classes. Our goal is that we use dark net/yolov3 in openCV to classify objects in Python.

An End-to-End GUI-Based Real-Time …

23

4 Materials and Methods The data consists of a group of shape images (m, 512, 512, 3). In addition to the known groups, the output is called a group of video frames. If cc is expanded into an 80-dimensional vector, each video frame is represented by 6 numbers (c by, bx, bw, bh, pc), resulting in 85 numbers for each bounding box. In YOLO V3, the prediction map is shown in such a way that any cell can guess exactly how many video frames are present. Consider the following example, (B ×(5 + C)) enters a map in the future. Each bounding box B is used to detect a particular type of object, and B shows how many frames are there that every cell can guess. Each video frame B has 5 + C attributes, which represent the dimensions, center coordinates, objectness ranking, and C trust levels across every class video frame. For each cell, YOLOV3 predicts three video frames. Since the center of the object falls in that cell’s receptive region, each cell in the feature map should be able to predict an entity using many of its video frames. This will be achieved based on how YOLO is trained to detect any given entity as a one video frame. This will be done on how YOLO is educated, responsible for detecting any given object is just one video frame [16]. Next, we have to figure out which of the cell to which this video frame belongs. The next step is to determine which cell this bounding box belongs to. To accomplish this, the input picture is divided into a grid with equal number of dimensions for the final result. The dimensions of the feature map will be 14 × 14.When the cell containing the center of an object’s ground truth on the input image is chosen, the object is expected. The cell that holds the middle of the truth-telling box has been colored yellow in the image below. The red colored cell is now the seventh cell on the grid in seventh row, and we have assigned the same spot on the future map, which is the seventh cell on the future map in the seventh row, to the dog’s detection [17–23] (Figs. 1, 2, and 3).

5 Results A entirely different strategy is used by YOLO. For doing target recognition in real time, the algorithm first merges the entire image into a SNN. Then, it segments it into areas, predicting video frames and options within each. The projected possibilities are used to measure these bounding boxes. For each region’s possibilities, the projected possibilities are used to measure these bounding boxes. The best benefit of using YOLO is its outstanding speed, which is amazingly simple and can handle forty five frames per second. YOLO is also familiar with fallacious representations of objects. We will be using an openCV and Python, and in this, the object will be detected in a still picture the applying one of the accurate algorithms called YOLO. We will be

24

Fig. 1 Proposed YOLO v3 architecture for the implementation Fig. 2 Inputting an image attributes to the architecture

M. V. D. Prasad et al.

An End-to-End GUI-Based Real-Time …

25

Fig. 3 Flowchart showing the complete recognition process

using an openCV which is a computer vision library or framework that will support our YOLOV3 algorithm (Tables 1 and 2). Table 1 Values of real-time spotter

Real-time spotter

Train

mAP

FPS

Fast YOLO

2008 + 2013

54.2

151.5

40 Hz DPM

2008

32.1

32.5

YOLO

2008 + 2013

66.1

42.5

26 Table 2 Values of less than real-time spotter

M. V. D. Prasad et al. Low real-time spotter

Train

mAP

FPS

Fastest DPM

2008

32.9

16.5

CNN

2008 + 2013

74.9

1

R-CNN

2008 + 2013

62.3

20.5

YOLO VGG-16

2008 + 2013

63.8

6.65

6 Conclusion Nowadays, most of the approaches for taking attendance is time consuming and also require manual work for users. So in order to overcome the above challenges, uses of face recognition in the system for keeping track of student attendance based on CNN and YOLO algorithms have been successfully demonstrated, which will mark the attendance by acquiring the images with help of a camera which will compare the faces of the students in the image with already enrolled faces in the database. Finally, a student will be given present or absent based on the acquired image. This system is time efficient, very simple to implement, and cost effective with no special vendor hardware and software necessary for implementation. In this face detection and identification technique, we have introduced YOLOV3 algorithm. As YOLOV3 is very accurate in object detection, it comes with its own ability of speed and accuracy.

References 1. Kishore PV, Rao GA, Kumar EK, Kumar MT, Kumar DA (2018) Selfie sign language recognition with convolutional neural networks. Int J Intell Syst Appl 10(10):63–71 2. Kishore PV, Kumar KV, Kiran Kumar E, Sastry AS, Teja Kiran M, Anil Kumar D, Prasad MV (2018) Indian classical dance action identification and classification with convolutional neural networks. In: Advances in multimedia, vol 2018 3. Kishore PVV, Kumar DA, Sastry ASCS, Kumar EK (2018) Motionlets matching with adaptive kernels for 3-d Indian sign language recognition. IEEE Sens J 18(8):3327–3337 4. Prasad MVD, Lakshmamma BJ, Chandana AH, Komali K, Manoja MVN, Kumar PR, Prasad CR, Inthiyaz S, Kiran PS (2018) An efficient classification of flower images with convolutional neural networks. Int J Eng Technol 7(11):384–391 5. Prasad MVD, Jaya Sree G, Gnanendra K, Kishore PVV, Kumar DA (2006) Fire detection using computer vision models in surveillance videos 6. Maddala TKK, Kishore PVV, Kumar K, Kumar A (2019) YogaNet: 3D yoga asana recognition using joint angular displacement maps with ConvNets. IEEE Trans Multimedia 7. Prasad CR, Kishore PVV (2017) Performance of active contour models in train rolling stock part segmentation on high-speed video data. Cogent Eng 4(1):1279367 8. Kumar EK, Kishore PVV, Sastry ASCS, Kumar MTK, Kumar DA (2018) Training CNNs for 3-d sign language recognition with color texture coded joint angular displacement maps. IEEE Signal Process Lett 25(5):645–649 9. Prasad CR, Kishore PVV (2015) Morphological differential gradient active contours for rolling stock segmentation in train bogies. ARPN J Eng Appl Sci 11(5):2799–2804. ISSN 18196608 10. Talab MA, Awang S, Ansari MD (2020) A novel statistical feature analysis-based global and local method for face recognition. Int J Optics

An End-to-End GUI-Based Real-Time …

27

11. Koch G, Zemel R, Salakhutdinov R (2015) Siamese neural networks for one-shot image recognition. Department of Computer Science, University of Toronto 12. Rashid E, Ansari MD, Gunjan VK, Khan M (2020) Enhancement in teaching quality methodology by predicting attendance using machine learning technique. In: Modern approaches in machine learning and cognitive science: a walkthrough. Springer, Cham, pp 227–235 13. Ren S, He K, Girshick R, Sun J (2015) Faster r-cnn: towards real-time object detection with region proposal networks. In: Advances in neural information processing systems, pp 91–99 14. Kang K, Ouyang W, Li H, Wang X (2016) Object detection from video tubelets with convolutional neural networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 817–825 15. Ren S, He K, Girshick R, Sun J (2015) Faster r-cnn: towards real-time object detection with region proposal networks. In: Advances in neural information processing systems, pp 91–99 16. Prasad PS, Pathak R, Gunjan VK, Rao HR (2020) Deep learning based representation for face recognition. In: ICCCE 2019. Springer, Singapore, pp 419–424 17. Kumar EK, Kishore PVV, Kumar MTK, Kumar DA, Sastry ASCS (2018) Three-dimensional sign language recognition with angular velocity maps and connived feature resnet. IEEE Signal Process Lett 25(12):1860–1864 18. Prasad PS, Gunjan VK (2020, Dec) Feature descriptors for face recognition. In: 2020 IEEE 17th India council international conference (INDICON). IEEE, pp 1–4 19. Sallawar N, Yende S, Padgilwar V, Kale V, Gorlewar P, Varma G (2017) Automatic attendance system by using face recognition. Int Res J Eng Technol (IRJET) 4(4), April 2017 20. Arsenovic M, Sladojevic S, Anderla A, Stefanovic D (2017, Sept) FaceTime—deep learning based face recognition attendance system. In: 2017 IEEE 15th international symposium on intelligent systems and informatics (SISY). IEEE, pp 53–58 21. Lukas S, Mitra AR, Desanti RI, Krisnadi D (2016, Oct) Student attendance system in classroom using face recognition technique. In: 2016 International conference on information and communication technology convergence (ICTC). IEEE, pp 1032–1035 22. Chen TY, Chen CH, Wang DJ, Kuo YL (2010, Dec) A people counting system based on facedetection. In: 2010 Fourth international conference on genetic and evolutionary computing. IEEE, pp 699–702 23. Gaddam DKR, Ansari MD, Vuppala S, Gunjan VK, Sati MM (2022) Human facial emotion detection using deep learning. In: ICDSMLA 2020. Springer, Singapore, pp 1417–1427 24. Taigman Y, Yang M, Ranzato MA, Wolf L (2015) Deep face: closing the gap to human-level performance in face recognition. In: 2015 IEEE conference on computer vision and pattern recognition (CVPR), September 2014

A Multifactor Security Protocol for Wireless Payment-Secure Web Authentication Using Mobile Devices Challa Venkata Pranith, Valiveti Lohya Sujith, Kolli Sai Kiran, Pulivarthi Goutham, and K. V. D. Kiran

1 Introduction As registering become imminent, individuals are progressively relying on the Internet to use their enterprise over the Internet. The Web is generally a preferable choice for online services such as Internet enterprise, e-casting a ballot, e-banking egovernment, and soon something close to that … Online applications need a defensive tackle highlight to ensure confidential customer information. Multifactor authentication makes more and shifted dividers to shut out some unacceptable individuals from seeing your data.

1.1 Types of Authentication Systems Authentication using a single factor (SFA) (Fig. 1): The most straightforward sort of confirmation strategy is single-factor authentication. A individual matches one credential with the SFA to verify himself or herself online. A password to a username will be the best known example of this. Today, most verification utilizes this kind of verification methodology. Problems with single-factor authentication are as follows: • • • •

Vulnerability problems The lacking of a backup stronger authentication The login credentials need to be strong enough We should not login in multiple devices as we cannot secure them. Authentication using two factors (2FA) (Fig. 2):

C. V. Pranith · V. L. Sujith · K. S. Kiran · P. Goutham · K. V. D. Kiran (B) Koneru Lakshmaiah Education Foundation (Deemed to be University), Vaddeswaram, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_4

29

30

C. V. Pranith et al.

Fig. 1 SF authentication

User Enters Username & Password

User Enters Mobile Number

User Receives One Time Password

User is Authenticated

User Enters OTP & Authentication is done

Fig. 2 TF authentication

Two-factor authentication consists of single-factor authentication and a other factor of authentication, such as a mobile device, using only something he or she owns. Getting straight, 2 factors are utilized to validate an identity. Problems with two-factor authentication are as follows: • Phishing attacks or attempts • Attacking through social engineering or brute-force method. Authentication using multiple factors (MFA) (Fig. 3): This is basically the next level security authentication system where we can authenticate the user or verify the user in multiple factors of security in order to achieve enhanced security system.

What You Know -UN & PWD Fig. 3 MF authentication

Something You Own - Phone

Something You Are - Fingerprint

A Multifactor Security Protocol for Wireless …

31

2 Literature Survey In 2016, Ashok Nath [1] faced problems and difficulties in the two-factor authentication method as the “store-front” looks genuine, and the user ends up entering directly into the hands of an identify information which are very much to be secured, i.e., not to be exposed to the outside world other than ourself. Coming into other type of security violation is if the would-be hacker already has access to the device itself [2]. Then, the intruder tries to piggyback the transmission and either execute fraudulent transactions or access protected transactions when the user accesses either company or Internet services [3]. Kumar Abhishek and Prabhat Kumar [4] published a study on multifactor authentication systems in 2017 as new authentication challenges and opportunities ensure that a customer is who they claim to be. As the authentication process includes more considerations, the appreciation of authenticity increases exponentially [5]. In 2017, Rajeev Ranjan [1] give the preferred use and interpretation of multifactor authentication that current systems can use. It is referred to as strong authentication or multifactor authentication when the protection infrastructure uses two or more separate and different forms of authentication mechanisms to protect the well-being of legitimate multifactor authentication use combinations of anything to provide more remote authentication than usual, weak single-factor username and password “In 2018, Aleksander Ometov, Sergey Bezzateev, Sergey Andreev [6] proposed the challenges faced by MFA management as the identification of users is a key element for a cautious and systematic approach to the adoption and implementation of MFA solutions, where most problems arise from opportunities and potential problems [7].

3 Existing Protocol/System Protection is an important factor in the online payment system focused on the Web. There are numerous network vulnerabilities affecting Web protection arrangements and growing electronic sharing hazards [8–10]. The majority and first authentication process is nothing but password credentials, so we feel that this type of authentication/verification framework does not really verify or validate the character of the customers that the person in question claims to consistently meet the current electronic administrations requires a username and secret key are done to verify the character of the client [11, 12]. In regard to communications, classified information protection has consistently been an important issue. With equipment advancements that give customers the advantage of transparency used in mobile phones, individuals are currently investing more and more energy in these gadgets. In addition, with the viral popularity of Web-based media applications and single sign-on, customers generally do not escape varying potential risks with their data [13, 14]. The current m-payment schemes can be categorized into three key groups, as discussed in

32

C. V. Pranith et al.

• Payment mechanisms focused on accounts, • Payment system for mobile POS, • E-cash or E-wallets. Therefore, in order to improve more and more reliable protection for online payments, we have proposed a new multifactor authentication protocol. Currently, we have multiple factor authentication system in some of the online systems, but here the issue is with the non-MF authentication system for which we cannot have a next factor to authenticate or to verify the user or client, so in place of that physical factor authentication system or to enhance the current system, we are here with TIC (transaction identification code)-based authentication system which will act as a MF authentication system.

4 Designed System For high-hazard exchanges that expect admittance to customer data or the exchange of assets to different gatherings, single-factor authentication is deficient. Multifactor authentication methods need to be used for securing online transactions. We use multiple factors to authentication in our scheme of design, using two different modes [15, 16]. TIC and SMS are used to execute the execution. Although in previous approaches to the issue, SMS was used. We are incorporating the latest definition of TIC which is the designed method of authenticating a online transaction and user to previous approaches to the problem. TIC codes are used to authenticate the client or user treating as a next level perception to authenticate the user/client [17, 18]. Codes for TICs are as follows: • Disseminated to the consumer by a banking institution. • 8 bits of randomly provoked codes are accredited to the client/user. • Sophisticated sequences of digits or combinations of binary or mixture of numbers, characters, and special symbols may be feasible. • The generated TIC will be used for a single transaction itself.

4.1 Protocol Here, the below screenshot is the sample execution of our proposed TIC-based multifactor authentication as a Java enterprise application and our proposed or designed protocol is segregated into parts of flow for execution, which is mentioned below.

A Multifactor Security Protocol for Wireless …

4.1.1

33

Protocol Workflow

See Fig. 4. 1.

Initially, the client will receive a list of TICs along with the login credentials, respectively, from their banking institution. Each customer has only one login credential in database record; anyway for each online trade, the TIC

1. User gets User-Id, Password & List Of TIC from Bank 2. Login on Bank Website by Credentials 4. Login Authentication Success 5. Select the Mode Of Payment

User With Mobile Phone / PDA with Encryption/ Decryption Module

6. Fill the Details Of Payment With Merchant (Account Information to Transfer)

7. Send TIC to Authenticate Transaction

User Bank/ Financial Institution web server with Encryption/ Decryption Module

9. ACK From Bank 10. SMS to Verify Transaction 11. Confirm or Denied Transaction By Replying SMS 12. ACK Message From Server to Commit or Deny Transaction

Fig. 4 Designed protocol for multifactor authentication

3. Verify User Info

8 Bank/ Financial Institution Authorization Server

34

2. 3.

4. 5.

6. 7.

8.

9. 10. 11. 12.

C. V. Pranith et al.

code is uncommon. Customers can obtain a once-over of TIC codes as demonstrated by their subtitles from the bank authority or affirmed money-related establishment. Basic authentication of a Web-based user-id/password is to validate the clientuser by the authentication server/system. A bank verification worker can approve the username and secret key. The client will get a decision screen after client affirmation to continue to promote to the next step. After successfully logged in, user will be prompted with a greeting quote. A session key is also produced by this phase. So now here, the client/user has a choice of transactional modes, i.e., two payment methods were considered: the system based on credit cards and the electronic transfer based on accounts. Adding other modes to our device is straightforward. Here, the user will enter the requisite particulars of payment-related info. By basically selecting a TICs from the dropdown of TICs, the client can embed a TIC codes. Note that TIC codes are put away with nearby encryption on the cell phone, with secret key assurance. The worker for bank approval decodes the message got and extricates the TIC codes. It at that point tests the TIC codes got from the client by contrasting it and the TIC codes list put away on the data set worker in the client account data. It drops the pre-owned TIC codes from its information base if the two TIC codes coordinate and goes to the following stage. In the event that no TIC codes coordinate those in the information base, the confirmation worker denies the exchange to the client and presents a blunder message to the client. The bank server creates a user recognition that allows the user to log out of the Web portal free of charge. After the successful update for the current carrying transaction is completed, the authentication system/server will initiate the email to client-user about it. Here now the at the user end, through the smart gadget or PDAs, the user will prove himself by the way of SMS/email type of authentication. The server will notify the user of the status of the transaction.

Now division/segregation of the protocol into 4 parts which also include two-way authentication. Part 1: User Authentication to the Banking Institute Authentication Server See Fig. 5. Part II: Merchant Authentication to the Consumer See Fig. 6. Part III: User Authentication to the Banking Authentication Server See Fig. 7.

A Multifactor Security Protocol for Wireless …

35

1. Visit Merchant Website & Select Goods to Purchase & Choose Payment Option 2. Invoice Details & Banking Info with Certificate info of Merchant Services 3.http Request To Bank Website 4.Display Secure Login Page 5.Login Using User Id & Secret Password

6. Forward User Data For Authentication 7.Authentication Success generate Session Key

8. Welcome Message With Secret Session Key

Fig. 5 User authentication to the banking institute authentication server

9. Send Details Of Merchant Authentication With Merchant Certificate With Merchant Banking Info

11.ACK From MB (Positive/Negative ACK)

12. Merchant Authorization Results (Positive/Negative ACK) CA

CB

Fig. 6 Merchant authentication to the consumer

CBAS

MB

MA

36

C. V. Pranith et al.

13.Mode Of Payment 14. Display entry form Transaction

simple for

15.Fill the Entry Form by inserting Transaction Details 16.Send TIC to Authentication in Encrypted Format 17. TIC Authentication

19. Send ACK of Transaction with Message

18. TIC Authentication Success

Fig. 7 User authentication to the banking authentication server

Part IV: User Authentication by Email confirmation See Fig. 8.

5 Security The attack of phishing has been a common vulnerability technique. Attackers tactically capture users’ essential data and carry out the illicit transfer of funds. We consider different cases below to grasp the origin of the security. Case I: If attacker acquires login credentials: Confidential TIC codes issued to legitimate users and TICs are not available in public. For each online transaction, it is a one-time code and it is created randomly so attacker cannot guess the TIC. Case II: Transmitting TICs through a transmission channel which is not secured: The transmission of TICs is carried out by encryption, so it cannot be easily decrypted by attackers to access private user information on the server side. In addition, only once is one TIC used and then discarded. Case III: If attacker acquires login credentials along with the one of the TIC codes:

A Multifactor Security Protocol for Wireless …

37

20. Send SMS for Confirmation of Transaction

21.Reply the SMS with option YES/NO

22. If Reply “YES” send Notification of Payment and commit Transaction 23.Notificati on of Payment with Invoice Details

25.If Reply is “NO” Rollback the Transaction and send Notification of Cancellation

CA

CB

CBAS

MB

MA

Fig. 8 User authentication by email confirmation

If the attacker receives a sample of TICs from any phishing technique, the attacker will not be able to crack the next coming or next to be generated TICs as it is a random generated TIC for each transaction.

6 Conclusion The current validation protocol for online framework implementation is not to ensure that customers are safe from wholesale fraud, as the effect is that any aggressor gains entry into confidential customer data such as Mastercard number or record secret key and allows illegal shop trade that will be charged to the record of the significant customer for remote installation of the system, and we zeroed in on an application-layer security response to conduct a start to finish verification and privacy of information between remote client and Java-based protected worker. The introduced work suggested another Web client validation convention focused on a multifaceted verification method that is fully safe and easy to execute. Future work will try to work on creating a modern and reliable way for producing TICs. Installing the TICs on mobile of the user must also be an. The simple job is to discourage frequent customer visits to the banking institutions.

38

C. V. Pranith et al.

References 1. Article on security threats of mobile phones. http://news.zdnet.com/2100-1009_22-5602919. html 2. Seah W Dr, Pilakkat S, Jaya Shankar P, Tan SK, Kee CS, Roy AG, Ng E (2001) The future mobile payments infrastructure a common platform for secure m-payments. A joint study by institute for communications research and systems, December 2001 3. Nath A, Mondal T (2016) Issues and challenges in two factor authentication algorithms. In: Proceedings of the international journal of latest trends in engineering and technology, January 2016 4. Adi W, Mabrouk A, Al-Qayedi A, Zahro A (2004) Combined web/mobile authentication for secure web access control. In: Proceedings of the IEEE conference held on communications and networking, pp 677–681, March 2004 5. Adi W, Mabrouk A, Al-Qayedi A, Zahro A (2004) Combined web/mobile authentication for secure web access control. In: Wireless communications and networking conference, IEEE Communications Society, Atlanta, GA USA, vol 2, pp 677–681, March 2004 6. Merugu S, Reddy MCS, Goyal E, Piplani L (2019) Text message classification using supervised machine learning algorithms. In: Kumar A, Mozar S (eds) ICCCE 2018. Lecture notes in electrical engineering, vol 500. Springer, Singapore, ISSN 1876-1100 7. Pu Q (2010) An improved two-factor authentication protocol. In: Proceedings of 2010 second international conference on multimedia and information technology, 24–25 April 2010 8. Shaik AS, Karsh RK, Suresh M, Gunjan VK (2020) LWT-DCT based image hashing for tampering localization via blind geometric correction. In: Kumar A, Senatore S, Gunjan VK (eds) ICDSMLA 2020. Lecture notes in electrical engineering, vol 783. Springer, Singapore 9. Halonen T (2002) A system for secure mobile payment transactions. Supervisor: Professor Teemupekka Virtanen, Helsinki University of Technology, Department of Computer Science and Engineering, January 2002 10. Shaik AS, Usha S (2019) Sensor based garbage disposal system. Int J Inno Technol Expl Eng (IJITEE) 8(4S2):164–167. ISSN: 2278-3075 11. Tiwari A, Sanyal S, Abraham A, Sanyal S, Knapskog S (2007) A multi-factor security protocol for the wireless payment-secure web authentication using mobile devices, held on February 2007, IADIS International Conference, Applied Computing 2007, Salamanca, Spain 12. Saxena S, Vyas S, Kumar BS, Gupta S (2019) Survey on online electronic payments security. In: Proceedings of the 2019 Amity international conference on artificial intelligence (AICAI), 4–6 Feb 2019 13. Devadasu G, Sushama M (2016) A novel multiple fault identification with fast Fourier transform analysis, In: 1st international conference on emerging trends in engineering, technology and science, ICETETS, 2016 14. Gao J, Cai J, Patel K, Shim S (2005) Wireless payment. In: Proceedings of the second international conference on embedded software and systems (ICESS’05), Xian, China 15. Premalatha B, Babu PR, Srikanth G (2021) Compact fifth iteration fractal antenna for UWB applications. Radioelectron Commun Syst 64(6):325–329 16. Narayana VA, Premchand P, Govardhan A (2009) A novel and efficient approach for near duplicate page detection in web crawling. In: 2009 IEEE international advance computing conference, IACC 2009 17. Saba L, Sanagala SS, Gupta SK, Koppula VK, Johri AM, Sharma AM, Kolluri R, Bhatt DL, Nicolaides A, Suri JS (2021) Ultrasound-based internal carotid artery plaque characterization using deep learning paradigm on a supercomputer: a cardiovascular disease/stroke risk assessment system. Int J Cardiovasc Imag 37(5):1511–1528 18. Dash CSK, Behera AK, Nayak SC, Dehuri S (2021) QORA-ANN: Quasi opposition based Rao algorithm and artificial neural network for cryptocurrency prediction. In: 2021 6th International conference for convergence in technology, I2CT 2021

IoT-Based Greenhouse Monitoring R. Agilesh Saravanan, Gowri Priya, Sai Nishanth, Praveen Sai, and Vasanth Kumar

1 Introduction The greenhouse is the location where we can grow fruit and vegetables. Every plant or crop has a specific environment to grow. Nowadays, the environment, temperature, humidity are changing every day. As a result, plants do not grow well, and few crops end without any benefit. I think it is the main reason to lose a crop by farmers. As a result, this project can reduce damage to crops/plants and help plants to live for a long time. In this project, we observe the surrounding environment of plants and crops like temperature, humidity, light intensity, and soil moist by using DHT11, LDR, and soil moisture sensors. Here, we use a DC fan to cool down the temperature, light bulbs, and a water pump to wet the soil. Here, we add a Wi-Fi module ESP8266 to monitor parameters and notification when the fan or blub or water motor is switched on/off. Temperature, humidity, soil moisture, and light intensity will be collected by the project till the end of the project. In today’s greenhouses, the monitoring of parameters is important for the good quality and productivity of plants. However, certain parameters such as temperature, humidity, soil moisture, light intensity are required for higher plant growth in order to achieve the stated consequence. Thus, NodeMCU has been developed primarily as a greenhouse control unit using sensors. The NodeMCU microcontroller is used for this project. NodeMCU can receive feedback from a wide range of sensors and can control generators, lights, and various actuators. Few sensors are used to measure few parameters. DHT11 sensor is used to measure temperature and humidity values. The soil moisture sensor tests the water content of the soil. The LDR sensor is used at the modest depth. The exhaust fan, water pump, and artificial light are also connected to the NodeMCU. R. Agilesh Saravanan (B) · G. Priya · S. Nishanth · P. Sai · V. Kumar Department of Electronics and Communication Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, Andhra Pradesh, India e-mail: [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_5

39

40

R. Agilesh Saravanan et al.

All environmental parameters are dispatched to Android cell phone via offline. A cloud is used to ship environmental parameters to server. As a result, when less than 50% of the soil loses its moisture, the motor pump will immediately turn on to sprinkle the water and start sprinkling the water until the moisture drops to 55%, and the pump is turned off after that. At a given time, sensor information will be sent to the ThingSpeak server so that it can be monitored from anywhere in the world. ThingSpeak enables instant viewing of the data that the computers have transmitted to ThingSpeak. With the power to execute MATLAB programming in ThingSpeak, you will do online data interpretation and processing because it comes in. ThingSpeak is additionally used for prototyping and proof of IoT concept applications involving analytics. When the temperature is raised to or greater than 40 degrees, the fan is switched on, and when the temperature is a smaller amount than 28 degrees, the fan is transitioned. Similarly, where the light level is less than the normal amount, the electric lights are switched on automatically and off when there is enough sunlight. So, a person can track the parameters with an Android phone. This device is very useful for farmers to observe and monitor environmental parameters in their fields. Farmers do not need to head to their farms. Any variance within the environmental criteria may result in monetary losses in the agricultural and pharmaceutical industries and may pose a life-threatening danger to consumers of biomedical industries. These losses can be stopped by managing them instantly.

2 Literature Review In [1], the project is about a system using GSM and Ethernet, which reduces the power consumption, maintenance, and complexity. That project can be used in agricultural field, in nursery, and in botanical garden. Used technologies are GSM module, Ethernet. Easy to communicate and monitor is one of the advantages. System costs high is a drawback. In [2], a four-layer device architecture was developed with outstanding motion control functions using mobile acquisition. Layers used are perceptual (physical) layer, control layer, transmission layer, application layer. Raspberry Pi and Arduino chips were combined to work as data server. Due to compact size, Raspberry Pi and sensors were integrated into mobile system. Cycle redundancy check (CRC) checking was used to reduce data loss at transmission layer. Advantage of this project is it monitor the highest and lowest values at a given point in time. In [3], Authors said that “Some complications arose during field experiments. Because of the EMI power source, some incorrect records were recorded. Salt deposition and thus incorrect measurements were observed on the soil wetness sensor probes. The total SMS loss was 0.5%. According to Tseng et al., this amount of SMS loss is appropriate (2006). The missing records were not consecutive and were insignificant on the remote server (1.8%)”. For [4], authors had designed monitoring of greenhouses and a framework focused on MSP430. The CC2530 module is introduced in this article. The method has the

IoT-Based Greenhouse Monitoring

41

benefits of stable service, low power consumption, durability, and convenient usage, etc., thus meeting a great deal with modern requirement for greenhouse production. The [5] is a survey report on agricultural problems. Farmers, guy. Many tools have been used to boost words of the farmers. This paper reflects on a major irrigation crisis. There are a lot of irrigation schemes. They are automated. The study allows us to learn about different IoTs and GSM-related methods used in the irrigation system. This is the paper that contains a section on the wise explanation of the previous work. So that, we will hear more about the problems of agriculture. This report is helpful to learn about the developments in agriculture. Ten years to enable a detailed survey is to be undertaken to help farmers. In [6], the project aims at the greenhouse in facility agriculture and the developed intelligent knowledge monitoring system, and the conclusion can be made as follows: (i) (ii) (iii)

The star network control structure was constructed. Used ZigBee to set up wireless sensor network. A kind of cooperative control method based on time control, manual control, automatic control, intelligent control, and remote control was proposed.

3 Design of Greenhouse Monitoring The method for monitoring and controlling the green house is focused on the calculation of the light intensity, soil moist, temperature, and humidity of the sensor situated at the locations. The result can be seen in ThingSpeak.

3.1 Block Diagram Block diagram of IoT-based greenhouse monitoring is shown in Fig. 1. The change in environment is sent to NodeMCU by different sensors such as soil moisture sensor, LDR, DHT11, and it analyzes the data and sends commands to respective

Soil Moisture

LDR

Water Motor

Node MCU

DHT11

Artificial Light

Exhaust Fan

ESP8266

Fig. 1 Block diagram of greenhouse monitoring

Cloud (ThingSpeak)

42

R. Agilesh Saravanan et al.

controlled devices such as water motor, artificial light, and exhaust fan. This data is uploaded in cloud-based platform called ThingSpeak using in-built Wi-Fi module called NodeMCU.

3.2 Hardware Description Design of greenhouse monitor hardware is to monitor all the data that come from various sensors which have different parameters like light intensity, soil moisture, humidity, and temperature. Every sensor senses the change in their respective domain and transfers this change to the microcontroller nothing but NodeMCU which we use in this project. For example, soil moisture sensor senses the moisture level in the soil and transfers the information to the microcontroller. As NodeMCU itself consists of Wi-Fi module in it can be used to upload all the data to the ThingSpeak which is a cloud-based platform.

3.3 Software Description The software is designed to process the humidity, temperature, light intensity, soil moist value from sensor to NodeMCU microcontroller. Then, continue to monitor the parameters from microcontroller. The microcontroller NodeMCU is to convert analog to digital and send the value of sensor through serial communication to computer. Control the water motor, artificial light, exhaust fan according to the parameter’s values. It is very convenient to schedule part of this project. The DHT library is used in this software to read the humidity and temperature sensor (DHT11 basic) from the humidity and temperature sensor, which can be monitored using ThingSpeak.

4 Methodology The key purpose of this research work is to focus on monitoring the greenhouse and acting on soil moist, air, and lighting services. The components used are as follows: • • • • •

NodeMCU ESP8266 Wi-Fi module Soil moisture sensor LDR DHT11

IoT-Based Greenhouse Monitoring

43

NodeMCU is an open-source IoT platform. It includes firmware running on the Espressif Systems ESP8266 Wi-Fi SoC and hardware based on the ESP-12 module. Various sensors used here are the temperature and humidity sensor (DHT11), soil moisture sensor, LDR. The exhaust fans, the water pump, and the artificial light are the end equipment. In the event that the dirt dampness esteem is underneath the expressed level, the motor turns on and siphons the water through the tubing [7]. It very well may be performed consequently and constantly until the characterized edge esteem has been met. Here, all limit esteems are taken from the perception of the rancher. LDR represents light-dependent resistor. It is utilized to decide the power of the light inside the nursery. In the event that the sunshine level is like the sting, the LED activates consequently and off within the event that it is not equivalent. DHT11 screen is for temperature and mugginess detecting. In the event that the temperature esteem is higher than or equivalent to the limit esteem, the fumes fan turns on consequently and off in the event that it is not exactly the edge esteem. In the surrounding environment that the dampness level is not exactly or equivalent the threshold level defined in the DHT11 sensor, the fan turns on consequently and draws the ventilated of the nursery and it is switch off when that sensor value is more than the defined threshold. The suggested method fits well and has shown effective outcomes. End equipment such as light source, water pump, and fans within the greenhouse have been triggered according to the threshold conditions of parameters such as temperature, humidity, soil moisture values. The data obtained from the NodeMCU is sent to the ThingSpeak server, and the data is reflected on the server. Here, the data is read from the different sensors that the sensed data is sent to the ThingSpeak IoT network through the internet using the ESP8266 module installed on the unit [8]. The framework portrays the assortment of exercises completed by the various modules, for example, temperature, humidity, LDR, and soil wetness. It tests whether the estimations of the boundaries are underneath or over the edge esteems [9]. The actuators are set off and deactivated considering the present situation. ThingSpeak is an IoT platform that helps to capture, interpret, and visualize data. It has TCP/IP to send and receive data. That is why we build channels in ThingSpeak. The sensed data is seen in the generated channels. Here, we use ThingSpeak as a cloud platform.

44

R. Agilesh Saravanan et al.

Fig. 2 Sensors data shown on the ThingSpeak dashboard

5 Result Finally, we have created a webpage in ThingSpeak. We gathered all sensors data and uploaded in ThingSpeak website using NodeMCU, and it analyzes the data and represents it in graphical format as shown in Fig. 2. Here are some of the data like temperature, humidity, soil moisture, and light intensity which are gathered from sensors and displayed in ThingSpeak. Here, the data may be varied from time to time so that we get ups and downs in the graphical representation of data.

6 Conclusion The key benefit of this project is that all tasks to be done by controlling devices such as exhaust fan, artificial lights, water motor and to monitor climatic conditions such as temperature, relative humidity, light intensity, and soil moisture levels in the greenhouse atmosphere are automated and do not require human intervention. We can also add GSM module to send SMS to the user’s phone to reduce the expense of the internet, and users can get alerts without internet access to their phone.

IoT-Based Greenhouse Monitoring

45

References 1. Vimal PV, Shivaprakasha KS (2017) IOT based greenhouse environment monitoring and controlling system using Arduino platform. In: 2017 International conference on intelligent computing, instrumentation and control technologies (ICICICT), Kannur, pp. 1514–1519. https://doi.org/ 10.1109/ICICICT1.2017.8342795 2. Geng X et al (2019) A mobile greenhouse environment monitoring system based on the internet of things. IEEE Access 7:135832–135844. https://doi.org/10.1109/ACCESS.2019.2941521 3. Chen P, Liu B (2010) Advancement of the CAN transport based intelligent greenhouse control system. In: 2010 international conference on computer application and system modelling V10– 631-V10–644 4. Azevedo J, Santos F, Rodrigues M, Aguiar L. Dozing ZigBee networks at the application layer. Wirel Sensor Syst IET 4(1):35–41 5. Guo W, Cheng H, Li R, Lü J, Zhang H (2010) Nursery monitoring system based on wireless sensor networks. Exchanges Chinese Soc Agr Mach 41(7) 6. Lihong Z, Lei S (2011) Estimation and control system of soil moisture of large greenhouse group based on twofold CAN bus. In: IEEE Conference Publications, pp 518–521 7. Hanggoro M, Putra A, Reynaldo R, Sari RF (2013) Green house monitoring and controlling using android mobile application. In: 2013 International conference on QiR, Yogyakarta, pp. 79–85. https://doi.org/10.1109/QiR.2013.6632541 8. Kumar KG, Rao KN (2019) Autonomous greenhouse using internet of things with ThingSpeak. 2019 Int Res J Eng Technol (IRJET), e-ISSN: 2395-0056, p-ISSN: 2395-0072 9. Xia L, Wenhui L, Yixin S (2017) Greenhouse monitoring system design based on MSP430 and king view. In: 2017 32nd youth academic annual conference of Chinese association of automation (YAC), Hefei, pp. 111–114. https://doi.org/10.1109/YAC.2017.7967388

Home Automation Using Telegram Bot Sree Vardhan Cheerla, V. V. N. Chakravarthy, K. KishoreBabu, and V. GopiRam

1 Introduction A Raspberry Pi is a charge card estimated PC which can be utilized for creating different applications. Internet of Things is an organization of gadgets, for example, electrical machines for network which empowers these gadgets to interface and trade information. This task speaks to an adaptable method to control gadgets. In this venture, orders for control gadgets, for example, “Turn light on” will be associated with Raspberry Pi, and as per it, the necessary cycle will work through Wi-Fi [1, 2]. It likewise gives security from outsider clients. It permits controlling number of home apparatuses all the while. Python is utilized as the fundamental programming language which is default, given by Raspberry Pi [3, 4]. It gives an intuitive and easy-to-use customer gadget and observed extremely simple. Prior days, the web has been broadly utilized for cycles, for example, riding site pages, data search, talking, download files [5, 6]. The progression innovations, checking administrations web give collaboration gadgets hardware. Framework is introduced in a few spots like emergency clinics, laboratories, banks, and different enterprises, which significantly diminishes the expense and time alongside keeping up security and comfort [7, 8]. Any actual boundaries from the true like temperature, dampness, pressure, and so on can be observed using sensors associated with ideal area on the planet. Applications associate with workers apparatuses and utilizing convention which is a machineto-machine convention particularly utilized for gadget-to-gadget correspondence [9, 10]. Informing application gives visiting administration furthermore, bot administration, by making another bot utilizing the message application machines visiting to the bot. Design various gadgets on the other hand. Extremely simple gadgets in view of adaptability are simply introduced [11, 12]. S. V. Cheerla (B) · V. V. N. Chakravarthy · K. KishoreBabu · V. GopiRam Department of ECE, Koneru Lakshmaiah Education Foundation, Vaddeswaram, Andra Pradesh, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_6

47

48

S. V. Cheerla et al.

2 Literature Survey 2.1 Chat Box Michael Yuan IBM developer works as IBM’s engineer manual for planning and creating bots. A pleasant review of stages, systems, and AI benefits that can be filled in as bot foundation. Twitter’s comprehensive list of AI-driven bots and assistants organized by function [13, 14].

2.2 Run Your Own Bot API Server Bot API source code is currently accessible at message bot-programming interface. Running own Bot API worker locally, boosting your bots presentation. Added the technique logout, which can be utilized to log out from the cloud Bot API worker prior to dispatching bot locally [15]. Log out the bot prior to running it locally; in any case, there is no assurance that the bot will get all updates. Added the technique close, which can be utilized to close the bot occurrence prior to moving it starting with one neighborhood worker, then onto the next [16].

3 Design Methodology Household appliances also calculate status of the sensors. The panel applications control the devices using the node.js. It is also connected to a chat application like Telegram. By using it, we can control the light and fan that are connected to it. It will give us the temperature, and when the temperature is high, it will automatically turn on the fan.

3.1 Architecture Figure 1 shows use of Wi-Fi sensor hubs, USB webcam, and Raspberry Pi with Raspbian Jessie working framework introduced on it and Android telephone with Android application Telegram. Various sensor hubs are associated with the Raspberry Pi by means of Wi-Fi channel. MQTT’s convention is utilized for correspondence. Sensor hub contains sensor module like DHT11 sensor, Pi camera sensor, gets signal from the sensor, and distributes message as needs to the Raspberry Pi which goes about as a worker with MQTT broker. Python content is written in the Raspberry Pi to peruse messages as gotten from the sensor hubs utilizing MQTT convention. A choice is then taken, and afterward, a message is shipped off from the Android

Home Automation Using Telegram Bot

49

Fig. 1 Architecture

telephone utilizing Android application Telegram, which is a non-benefit cloud-based texting administration.

4 Implementation 4.1 Telegram Bot Telegram is an immediately messaging utility that we use every day to speak with the own circle of relatives. The loose and open-supply nature of tele helped builders release a hard and fast of APIs used to expand bots. Bots are the programs that automate chores. By the use of this bot, it is far viable equipment everywhere. Improvement of this bot jogging on Raspberry Pi is linked to sensors inclusive of domestic family house equipment inclusive of fans, lights, etc. The bot gets person commands with the aid of using Telegram and responds to them in consequence. The commands are strings that they are programmed to reply to the person. Alongside photograph or message, framework likewise sends date that the consumer is probably knowledgeable approximately the hour of occurrence (at the off hazard that at the off hazard that he/she cannot get Telegram message proper away due to no organization or a few different reason). Telegram is a non-advantage cloud primarily based totally texting administration. Customers can get admission to from Android, iOS, Windows Phone, Windows NT, Macintosh OS, and Linux [9]. Utilizing these, customers can ship messages, photographs, recordings, files, sound, and a few different documents. At the factor while any motion is diagnosed with the aid of using the framework, it

50

S. V. Cheerla et al.

is going to seize the photograph/video making use of the digital camera and ship it to the message utility [3].

4.2 Raspberry Pi Camera The camera module may apply advanced pleasant video, simply stills/photographs. It is something, however, hard to apply for tenderfoots, but has bounty to deliver to the desk stepped forward customers in case you are hoping to increase your insight. There are hundreds of fashions online of people making use of it for time-slip by, slow-movement, and different video intelligence [7]. You can likewise make use of the libraries we % with the digital effects.

4.3 DHT11 Four-pin system would require a resistance to be placed among pin one and pin two. The three-pin modules will more commonly have this resistance blanketed which makes the wiring truly simpler. Thus, Fig. 2 indicates becoming a member of Pi with a piece of three-manner DuPont link. Various carriers might also additionally twine the module sticks distinctively so test the PCB markings to apprehend VCC (+), facts, and ground (−). Fig. 2 Interface of Pi with DHT11

Home Automation Using Telegram Bot

51

Fig. 3 1 x 2-channel 5V relay module

4.4 Two-channel Relay 1 × 2-channel 5 V relay module is a switch interface board; it thoroughly can be managed straightforwardly with the aid of using an extensive scope of microcontrollers, for example, Arduino, AVR, PIC, ARM, etc. It makes use of a lowdegree spark off manage sign (3.3-5VDC) to manipulate the switch. Setting off the switch works the usually open or generally close contacts. It is regularly applied in a programmed manage circuit. To lay it out plainly, it miles a programmed alternate to manipulate a high-modern circuit with a low-modern sign.5 V switch sign fact’s voltage range, 0−5 V. VCC potential to the framework. JD-VCC hand-off with inside the pressure flexibly. JD-VCC and VCC may be a shorted (Fig. 3).

4.5 DC Fan Immediate flow (DC) of an engine could be a kind of an electrical energy to energy.

5 Software Tools 5.1 Telepot and Data Publishing in Telegram App Telepot encourages you to fabricate applications for a wire larva API. It deals with Python pair of and the Python three. The related outcome shows the detector info that has gotten from the temperature and viscosity sensors and moreover the standing of the equipment; we will understand what proportion temperature may be calculable, controlled, and differed by given air conditions.

52

S. V. Cheerla et al.

6 Results and Discussions The end result in Fig. 4 indicates that usage instructions have interaction with Pi, Telegram app, and various sensors. At after giving commands, /getouts—indicates the repute of the 2 relays, /setout1 ON or OFF, /setout2 ON or OFF units, //gettemp—indicates the real temperature, and //gethum—indicates the real humidity. Table1 shows few display values, which have been plotted. /start-issued, it indicates that the technique is not begun out, and it indicates that command is not within side the listing. In the identical technique, we are able to join exclusive sensors to screen and manage the devices.

Fig. 4 Hardware setup

Home Automation Using Telegram Bot Table 1 Temperature and humidity values by DHT11

53

S. no.

Time

Temperature values (in Celsius)

Humidity (percentage)

1

10 s

300

64

2

20 s

32

65

3

30 s

35

67

4

40 s

27

45

5

50 s

25

44

6

60 s

15

35

7

1 min 10 sec

22

41

7 Conclusion The home framework tentatively incontestable figure assistance of the various techniques existent message application. Management apparatuses accomplished techniques distantly applications. This may assist consumer with examining within house whenever anyplace. Framework is like an attachment which might anyplace not most expense, however rather a lot of protection. Solitary conceivable utilizing instrumentality supply programming.

References 1. Alkar Z, Buhur U (2005) An internet based wireless home automation system for multifunctional devices. IEEE Trans Consum Electron 51(4):1169–1174 2. Zhao Y, Ye, Z (2008) A low cost GSM/GPRS based wireless home security system. IEEE Trans Consum Electron 54(2) 3. Zeng X, Fapojuwo AO, Robert J (2007) Design and performance evaluation of voice activated wireless home devices. IEEE Trans Consum Electron 52(31):983–989 4. Ahmed SM, Kovela B, Gunjan VK (2020) IoT based automatic plant watering system through soil moisture sensing—a technique to support. Adv Cybern, Cognit Mach Learn Commun Technol 643:259 5. Anvekar RG, Banakar RM (2017) IOT application development: home security system. In: IEEE technological innovations in ICT for agriculture and rural development (TIAR 2017), Apr 7–8, 2017, Chennai, India, pp 68–72 6. Rakesh VS, Sreesh PR, George SN (2012) An improved real-time surveillance system for home security system using BeagleBoard SBC, Zigbee and FTP webserver. IEEE Int Con 1240–1244 7. Gaddam DKR, Ansari MD, Vuppala S, Gunjan VK, Sati MM (2022) A performance comparison of optimization algorithms on a generated dataset. In: ICDSMLA 2020, pp 1407–1415. Springer, Singapore 8. Ansari AN, Sedky M, Sharma N, Tyagi A (2014) An internet of things approach for motion detection using Raspberry Pi. In: IEEE international conference intelligent computing and internet of things, pp 131–134 9. Yamanoor NS, Yamanoor S (2017) High quality, low cost education with the Raspberry Pi. In: 2017 IEEE global humanitarian technology conference (GHTC), San Jose, CA, 2017, pp 1–5 10. Zhu J, Huo L, Ansari MD, Ikbal MA (2021) Research on data security detection algorithm in IoT based on K-means. Scal Comput: Pract Exper 22(2):149–159

54

S. V. Cheerla et al.

11. Balakrishna S, Solanki VK, Gunjan VK, Thirumaran M (2019) A survey on semantic approaches for IoT data integration in smart cities. In: International conference on intelligent computing and communication technologies, pp 827–835. Springer, Singapore 12. Kumar SS, Rao GS, Voggu S (2021) Improvement of DC bus bar voltage for microgrid system using renewable energy sources. Lect Notes Netw Syst 609–616 13. Anjankar SC, Zalke J, Pandey SR, Misal N, Jawarkar P (2020) Vehicle monitoring system based on internet of things (IoT) for smart cities. Helix Scient Expl | Peer Rev Bimonthly Int J 10(01): 222–227. Retrieved from https://helixscientific.pub/index.php/home/article/view/87 14. Dhaya R, Kanthavel R, Mahalakshmi M (2021) Enriched recognition and monitoring algorithm for private cloud data centre. Soft Comput 15. Makled Esraa A, Halawa Hassan H, Daoud RM, Amer Hassanein H (2015) Wi-Fi-based hierarchical wireless networked control systems. Electronics and Communications Engineering Department American University in Cairo (AUC) Cairo Egypt IEEE 16. Saravanan NP, Thamilselvan R, Loheswaran K (2021) Prediction of neurological disorder using deep learning network. Oxidation Commun 44(1):171–187

Application of IoT in Hospital Management Sree Vardhan Cheerla, Syed Inthiyaz, V. Subba Reddy, K. SaiSaketh, N. Kiran Bhavya, and M. Vinay Kumar

1 Introduction 1.1 Internet of Things The Internet of Things (IoT) refers to a collection of general devices like electronic appliances, vehicles, and sensing devices like sensors, actuators which have the ability to transfer or receive data over the internet [1, 2]. The Internet of Things has a large part of its usage in general wards of hospitals. This increases the safety measures, efficiency of the work will be enhanced, and time-saving for staff is quiet beneficial for patients as well as hospitals for management (Fig. 1). Building blocks of IoT mainly rely on four things—sensors, processors, gateways, and applications. A useful IoT system is made up of huge number of nodes, where each node has got its own individual characteristics. Sensors are the so called Things of the system. They should move all the time in nature which says that they should be ready to receive the real-time data. Those sensors can be either autonomous in nature or the user-controlled ones.

1.2 Vascular Air Embolism Vascular air embolism is very rare but a dire situation. Gas molecules enter into veins when the saline bottle is empty. The gas molecules travel through the right heart to the pulmonary circulation which causes a negative pressure which can definitely cause the blood flow out of the body into saline bottles [4, 5]. This situation ultimately S. V. Cheerla (B) · S. Inthiyaz · V. Subba Reddy · K. SaiSaketh · N. Kiran Bhavya · M. Vinay Kumar Department of ECE, Koneru Lakshmaiah Education Foundation, Vaddeswaram, AP, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_7

55

56

S. V. Cheerla et al.

Fig. 1 Building blocks of IoT [3]

makes the vascular air embolism come into picture. These cases occur during hectic schedules when enough members are not there to supervise the wards. To overcome such situations, a holder is being made for the saline bottle by IoT and ultrasonic sensing devices to track down the level of saline in the holder [6]. This not only helps to avoid this kind of situation but also helps to do things as per the time and systematic manner. It also alerts the staff whenever the saline level is below the defined value.

1.3 Fire Accident Air quality and fire accidents are few major accidents noticed in hospitals which leaves a large amount of people in a dangerous situation and sometimes it leads to death. Quality and fire issues were being detected by gas sensor MQ-135 [4]. The alarm system designed with IoT notifies to switch on air purifiers or to open windows whenever the air quality drops below the threshold value.

1.4 Power Consumption One of the common problems that every private and public hospital faces is high electricity bills and irrelevant power consumption [7]. By using smart lights, fans, and air conditioners to narrow the usage of electricity will reduce power issues.

Application of IoT in Hospital Management

57

2 Literature Survey This literature survey approaches with the study of ideas which are used in the past in the implementation of smart hospitals using IoT. The implementation which is stated in the paper titled “IoT Solutions for Hospital” [1] spotted vascular air embolism, excessive power consumption, and air-quality issues. The way of the approach used in the reference paper to overcome issues was taken and modified by using accurate, sensitive sensors and to meet real-life conditions.

3 Block Diagram Figure 2 provides you the input and output components of the project. The sensors ultrasonic, PIR, MQ-135 were used as input, and the Blynk app, LED, and a buzzer were used for highlighting the output at the other end.

3.1 Components 3.1.1

MQ-135 Gas Sensor

SnO2 material which is present in MQ-135 gas sensor has got a high level of resistance for gas sensing in clear air. As the polluting gases concentration in the surroundings increases, the sensor’s resistance gradually decreases. MQ-135 sensor measures PPM by (Rs/Ro) ratio [8]. Harmful gases like ammonia, nitrogen oxide, alcohols, aromatic, and smoke and air quality were being detected by MQ-135 [8, 9]. It can be used as both an analog and digital sensor.

Ultrasonic Sensor

PIR Sensor

MQ135 Sensor Fig. 2 Block diagram of project

Blynk

ARDUNO UNO

LED

Buzzer

58

3.1.2

S. V. Cheerla et al.

PIR Sensor

It detects the changes in the infrared light across a certain distance and gives an electrical signal as output in response to the detected IR signal [6]. The sensor detects the IR rays emitting from a person when the person moves away or within the range of the PIR sensor. This sensor is chosen because of its dual or a pair of sensing the infrared signals by canceling the unwanted temperature variations that existed within the range which helps to detect IR signals from human beings. PIR sensor detects the objects within the particular range of 10 m approximately. The actual detecting range is 5–12 m [10].

3.1.3

Ultrasonic Sensor

It detects the objects by emitting sound waves and identifies the distances by detecting the reflected rays [11]. The sensor converts the reflected sound into an electrical signal. It uses piezoelectric crystals for transmitter and receiver. In this project, it has been used as a level detector to detect, monitor liquid levels. Being a saline level detector, it detects the saline fluid level in the holder; when it reaches the threshold value, it alerts the staff to change the saline bottle of the patient. Vascular air embolism is very rare but a direful situation. Gas molecules enter into veins when the saline bottle is empty [12]. As a level detector, it detects the saline fluid level in the bottle; when it reaches the threshold value, it alerts the staff to change the saline bottle of the patient.

3.1.4

Arduino UNO

Arduino is an ATmega328 microcontroller board which has got some pair of digital and analog input and output pins which have been used to interface with other boards or devices helpful in building applications [13, 14]. It is programmed with the Arduino IDE and powered by an external 9 V battery or by USB cable.

3.1.5

Blynk APP

Blynk began as a crowd-funding initiative that raised just over $50 k. The platform went live on May 22 and has been updated on a regular basis since then. In reality, new Twitter, push, and email integration gadgets are being added to the app. Most Arduino boards, the ESP8266, Raspberry Pi versions, Particle Core, and a few other popular microcontrollers are currently supported by Blynk, and new things are being added over time.

Application of IoT in Hospital Management

59

4 Design and Functionality 4.1 Saline Level Indicator In the process of expressing the saline level and preventing air embolism in the saline bottle, the ultrasonic sensor has been used [15]. As soon as the ultrasonic sensor distance from the saline increases, it indicates the lack of saline in the saline bottle. Hence, a gate value is set, and whenever the sensor crosses the gate value, it is made to notify the hospital staff [16, 17]. So, they can get to replace or refill the saline bottle before it gets completely empty.

4.2 Harmful Gas Detection A buzzer was interfaced with Arduino to indicate the harmful gas or smoke present in the surrounding environment. The MQ-135 gas sensor is more sensitive to smoke than the fire. Hence, we could understand that particles of the smoke react with the resistive material present in the sensor, by modifying the resistance of material (tin dioxide) [18, 19]. This can be used in hospitals for identifying the gas leakage much earlier.

4.3 Reduction of Power Consumption A PIR sensor is used in the power consumption reduction in the hospitals. The sensor’s functionality is tested by experimenting the PIR sensor with opaque bodies. The PIR sensor catches the object’s motion; whenever an individual enters the space, it activates the lights and it remains during a normally off state. The campaigns by utility providers of power, let us consider an example in Nigeria, on efficient energy management mean nothing to consumers, thereby over stressing the components and appliances at power system [10, 20]. This Fig. 3 presents the working principle of the project which monitors the motion, saline level, and the harmful gas or smoke in parallel. By the figure, you can come to conclusion that the sensors monitor the surroundings continuously.

5 Hardware Connections This circuit diagram in Fig. 4 provides the connections with the ultrasonic, PIR, MQ-135 sensors. As soon as the sensor reaches their required conditions, they send a notification to the users. The ultrasonic sensor sends a notification to the user’s

60

S. V. Cheerla et al.

St

1.Ultrasonic Sensor monitoring the distance from saline water

1.MQ135 Sensor checking for harmful gas in the surrounding

1.PIR Sensor checking that if any person enters the room

If If Saline Level
“Dashboards” to create a dashboard, click on the “dashboard” icon on the top left of the user interface, hold all default fields, and then create them. Choose on add new widget, you will be prompted with a bunch of knowledge visualization options, choose one of them, i will be able to use a line chart to plot graph using the uploaded received data from esp32Module. Top on add variables, then select the device, and choose the variable. Add a reputation for the graph. Keep the remainder of the choices default. Later create it. Since our device did not interact with Ubidots, it will show No Data Found. Once our device started posting data, we can see the info here itself.

108

A. Vamseekrishna et al.

12.

By using Arduino IDE, we will program the ESP32. So, we should always have the esp32 add on installed in our Arduino Ide. After successful setup of the specified credentials over the esp32 module, and making a necessary reference to Ad8232 and Arduino interfacing with esp32, we will see the specified heart pulses over the Ubidots platform.

13.

5 Experimental Measurements and Results Ubidots is used as a cloud server to check and manipulate the data from anywhere through Wi-Fi. As we have implemented the hardware in Arduino UNO board, in notebooks and smartphones, this device will work. In Fig. 10, laptop is used. The electrodes are placed on person; the heart monitoring is seen in Fig. 11. Fig. 10 Communicate with Ubidots

Fig. 11 Heart rate of the patient

Low-Cost ECG-Based Heart Monitoring System with Ubidots Platform

109

Fig. 12 Heart rate of the patient in Ubidots

The information is stored via Wi-Fi module in the Ubidots, which is linked to the hardware device. By the stored data, the graph is made so that the patient heart rate is monitored from anywhere through cloud service. The ECG signal is received in the laptop data by a USB cable attached directly to Arduino Uno by a Wi-Fi module. We stored the ECG signal data in the cloud. The Electrocardiogram signal is taken from a woman who is 40+ years old Figs. 10 and 12.

6 Cost Analysis See Table 2. Table 2 Cost of the equipment Name of the component

Quantity

Unit price (INR)

Total cost (INR)

Arduino UNO

1

250

250

Connecting wires

10

10

10

AD8232 ECG sensor

1

750

750

Bread board

1

50

50

Wi-Fi module

1

100

100

ECG electrode connector—3.5 mm

1

50

50

110

A. Vamseekrishna et al.

7 Conclusion This project is to develop and implement an Ubidot-based low-cost ECG and heart monitoring system. The desired ECG signal has been found and checked, and this is for a physician. The heart rate and ECG signal are visualized by our built scheme. We will add irregular ECG identification functionality to this scheme in our future work. The cost of this device is just INR, making it relatively cheap and suitable for developing and underdeveloped countries. For the doctors which is useful and any cardiac patients and in any other developing world, such as Bangladesh and Sri Lanka, this device is an excellent option.

References 1. Al-Busaidi AM, Khriji L (2013) Digitally filtered ECG signal using low cost microcontroller. In: International conference on control, decision and information technologies (CoDIT), pp 258–263 2. Nichols M, Townsend N, Scarborough P, Rayner M (2014) Cardiovascular disease in Europe 2014: epidemiological update. Eur Heart J 35(42):2950–2959 3. Ueshima H, Sekikawa A, Miura K, Turin TC, Takashima N, Kita NY, Okamura T (2008) Cardiovascular disease and risk factors in Asia a selected review. Circulation 118(25):2702– 2709 4. Khor GL (2001) Cardiovascular epidemiology in the Asia-Pacific region. Asia Pac J Clin Nutr 10(2):76–80 5. Liu B, Shi G, Zhao W (2017) The design of portable ECG health monitoring system. In: 2017 29th Chinese control and decision conference (CCDC) 6. Kamble A, Birajdar A (2019) Internet of things based portable ECG monitoring device for smart healthcare. In: 2019 fifth international conference on science technology engineering and mathematics (ICONSTEM) 7. Gunjan VK, Reddy MJ, Shaik F, Hymavathi V (2018) An effective user interface image processing model for classification of Brain MRI to provide prolific healthcare. Helix J 8(3):2129–2132 8. Kamble P, Birajdar A (2019) IoT based portable ECG monitoring device for smart healthcare. In: 2019 fifth international conference on science technology engineering and mathematics (ICONSTEM) 9. Shaik AS, Usha S (2019) Sensor based garbage disposal system. Int J Innov Technol Exploring Eng (IJITEE) 8(4S2):164–167. ISSN: 2278-3075 10. Satija U, Ramkumar B, Sabarimalai Manikandan M (2017) Real-time signal quality-aware ECG telemetry system for IoT-based health care monitoring. IEEE Internet Things J 4(3):815–823 11. Allam VK, Madhav BTP, Anilkumar T, Maloji S (2019) A novel reconfigurable bandpass filtering antenna for IoT communication applications. Prog Electromagnet Res 96:13–26 12. Shaik AS, Karsh RK, Suresh M, Gunjan VK (2020) LWT-DCT based image hashing for tampering localization via blind geometric correction. In: Kumar A, Senatore S, Gunjan VK (eds) ICDSMLA 2020. Lecture notes in electrical engineering, vol 783. Springer, Singapore 13. Gao Y, Soman VV, Lombardi JP, Rajbhandari PP, Dhakal TP, Wilson D, Jin Z (2019) Heart monitor using flexible capacitive ECG electrodes. IEEE Trans Instrum Meas

Low-Cost ECG-Based Heart Monitoring System with Ubidots Platform

111

14. Wannenburg J, Malekian R, Hancke GP (2018) Wireless capacitive-based ECG sensing for feature extraction and mobile health monitoring. IEEE Sensors J 18(14):6023–6032 15. Winter BB, Webster JG (1983) Driven-right-leg circuit design. IEEE Trans Biomed Eng 62–66 16. Vamseekrishna A, Madhav BTP (2018) Defected ground structure switchable notch band antenna for UWB applications. In: Smart computing and informatics. Springer, Singapore, pp 139–145

Design of Hybrid Soft Computing Techniques for Estimation of Suspended Sediment Yield in Krishna River, India Arvind Yadav, Sanjay Vishnoi, Pragati Mishra, Devendra Joshi, and Haripriya Mishra

1 Introduction Weathering influences and shows its effect on river sediment load, which is one of the most significant nature’s landscape transformation processes. The phenomena of weathering being enforced on different kinds of soils, rocks and their mobility in watersheds as well as rivers are considered under the complex hydrological and environmental problems. The suspended sediment load is nothing but sediments that pass through the rising elements of turbulent currents with the river water, stay suspended and stay stationary for a while [1]. Due to changes in the global atmosphere, the correct measurement of suspended sediments has drawn attention in water resources and environmental research over the last few decades. Sedimentation has an impact on dams’ physical strength; hence, the efficiency of dams gets reduced gradually based on the intensity of sediments. For all these reasons, the sediment yield estimation and measurement have become very much essential. During the times of cyclones and heavy rainfalls, the measurement of suspended sediment becomes the harder job as any other. The involvement of several complex processes in sediment yield causes inaccuracy in computation of sediments, when the traditional methods are used. For the estimation of sediment yield, multiple linear regression (MLR) (a classical statistical model) was used [2]. The major demerits of linear models include restricted ability in capturing the non-linearities in both hydrological and environmental data. Artificial intelligence methods such as the Artificial Neural Network (ANN) have therefore

A. Yadav (B) · S. Vishnoi · P. Mishra · D. Joshi Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, Andhra Pradesh, India e-mail: [email protected] H. Mishra Department of Civil Engineering, Gandhi Institute for Technology, Bhubaneswar, Odisha, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_13

113

114

A. Yadav et al.

been described as an effective method for precise estimation in pursuit of an alternative solution to the current mechanism and physically dependent sediment yield models. The main objective of ANN is to create learning methods through pattern recognition which can learn from information and predict the SSY [3]. The ANN has found reasonable artificial intelligence-based methods that deal with flexible mathematical structures and has the capability for identifying the relationship between inputs and output (sediment yield). The relationship between input and output can be very complicated and not linear. Various researchers have successfully applied artificial intelligence and machine learning methods for solving global nonlinear problems in water resources and other domains [4–15]. Broad studies on the use of artificial neural networks (ANNs) in suspended sediment estimation have been done [7, 16]. Inappropriate characteristic of trial and error methods throws challenges to the traditional algorithms and hence, in designing robust soft-computing models for the selection of ANN’s model parameters (neurons, weights, transfer functions, and learning parameters), is a very challenging task. In the last decade, researchers have been using the GA method as an alternative to address parameter selection limitations for ANN models [17, 18]. During a complex, adaptive system configuration creating the GA will produce a collection of optimal solution points. The GA in ANN modeling is very useful because of the potential to avoid trapping the ANN in the local minimum region and also because of the optimized parameters’ selection of the ANN model [8, 17]. The ANN model was developed in this paper to simultaneously optimize all the ANN parameters (transfer functions, hidden-layer neurons, weights in the network, and combinational coefficients) using the GA and apply them in the Krishna River basin to estimate sediment yield.

2 Study Area and Data Used The prediction of sediment load is done in the Krishna River, India. Based on the catchment area, Krishna River has the fourth rank. The overall catchment area covered by the Krishna River basin is 70,614 km2 , comprising 8% of India’s geographical area. In three states, Maharashtra (26.36%), Andhra Pradesh (29.81%), and Karnataka (43.8%) occupied the Krishna River basin area. The drainage area of 71,417 km2 is managed by the Krishna River gauge station at Waddepally. It falls from the Hamsaladeevi in Krishna district of Andhra Pradesh, India, at an altitude of 870 m above sea level. The total area of the Krishna river basin is approximately 8% of the total geographical area in India, that is 25,8948 km2 . The overall length and width of the Krishna River is between 701 and 672 km, ranging from 73°17′ to 81°9′ east longitude to 13°10′ to 19°22′ north latitude. Present sediment load data has shown that it ranks fourth among India’s peninsular rivers. In this study, hydro-climatic data is obtained from the Krishna River basin at Waddepally Gauging Station. The geographic location of Krishna basin with Waddepally gauge station has shown in Fig. 1. For this analysis, Krishna district was chosen as the study site with data

Design of Hybrid Soft Computing Techniques for Estimation …

115

Fig. 1 Site map of the Krishna River basin with its main stream and gauge station at Waddepally

availability (water discharge, water level and suspended sediment), primarily due to its strategic position. Highlight of this basin of Krishna is that the main portion (75.86%) of the basin is covered by farmland that provides food for most of southern India. Around 10% of the basin region contains forest, wastelands occupy about 7% of the total basin area, and water sources cover about 4% of the basin area. In the Krishna basin, there is a tropical atmosphere dominated by the southwest monsoon which precipitates the basin mainly. The Krishna basin gets its peak rainfall during the Southwest monsoon, as in most other regions of India. The present analysis utilizes 40 years of monthly water level, water release, and suspended sediment yield data for the creation of the proposed model from January 1966 to December 2005. Data is divided into training (70%), validation (15%), and testing. Before development of the model, data is normalized between the 0 and 1 for developed the efficient model. Detailed description about the normalization is demonstrated by various researchers [8, 17].

116

A. Yadav et al.

3 Methodology In order to obtain reliable performance, a combination of genetic algorithm (GA) and artificial neural networks (ANN) is used for estimation of sediment. The GA can solve many complicated nonlinear problems with the ANN. By correlating the data sets for input and output, the ANN is a robust and well-established statistical tool for achieving estimated performance. The ANN algorithm imitates the biological concept of the brain which is related to nervous system. The Multilayer perceptron (MLP)-based ANN with the back propagation training algorithm Levenberg–Marquardt (LM) has been used for estimation of sediment yield, whose key advantages are robustness and fast computation. Updating of the connection weight of the network and bias weights updation is done using this LM algorithm. Because of its speed of response, the robust MLP neural network model is developed with the Levenberg–Marquardt (FFBP-LM) feed forward and back propagation algorithm. Therefore, it is the well-established supervised training algorithm, although more memory is needed. Each layer is connected to other layer with connection weight and bias with the three distinct layers of the feed forward network, i.e., input layer, hidden layer, and output layer. By weighted interconnection link, which has substantial data in the ANN, the one-layer neurons are connected with those of a separate other layer neurons. Numerous studies have shown that one hidden layer is suitable for any nonlinear dynamic function and to minimize ANN’s structural complexity [19]. The MLP-based LM algorithms are described briefly in [8, 18]. Several factors, such as the input data, the quantity of neurons, the activation function, and the original weight values affect the performance of the MLP neural network models. The wrong collection of these variables could clearly represent the construction of a poor ANN model. All of these variables for ANN were picked simultaneously using the genetic algorithm in the present study. The genetic algorithm (GA) was eventually used to find the best ANN model parameters [20], which is an optimization algorithm based on the global population. It utilizes different genetic operators in the chromosomal population of individuals, such as mutation, selection, and crossover. In several research [21–23], the GA optimization approaches are widely used. Current research encourages use of hybridized GA with ANN represented as GA-ANN. The LM algorithm was used for artificial neural network training, and all the ANN parameters were chosen by GA. The selection of the simultaneous optimization of the ANN’s training parameters by using the GA has resulted in a much reliable optimum solution and no chance to trap in local optimal point.

4 Result and Discussion For developing the GA-ANN model, the MATLAB software with customized code was used. For the selection of optimum ANN model parameters, the normalized

Design of Hybrid Soft Computing Techniques for Estimation …

117

data was used. The model has given a sequence of final solutions at a maximum generation stopping conditions which is equal to 50. The fitness function value is given in Fig. 2 over different generations of the GA-ANN model. Along with the mean fitness value of 19.988, the highest fitness was considered to be 0.00565. In order to develop the robust GA-ANN model, the best chromosome of the best fitness function explains the need to pick all the input parameters, i.e., water discharge and water level. The findings have showed that the selected optimum number of nodes (neurons) is 30 in the hidden layer of the ANN. The tan-sigmoid function was picked as a transformation function in the hidden and output layer optimally. The optimized combination coefficient (µ) value was chosen by the GA-ANN model as 106 in the Levenberg–Marquardt algorithm model. A uniform crossover with a likelihood rate of 0.6 in the GA-ANN model was used in this analysis. Figure 3 demonstrates the variation in the performance of the GA-ANN model during the training, validation, and testing phases. For calculating the output of the proposed model, normal statistical error measures (mean square error—MSE, root mean square error—RMSE, mean absolute error— MAE, coefficient of correlation—r, and error variance—VAR) are used. The results of different models are compared according to the evaluation criterion. In this GAANN model, the generalization and the reliability of this model were checked using the data set of testing phase. In the validation, training, and testing stage, the statistical error of the proposed GA-ANN model was presented in Table 1. Table 1 shows that RMSE was very low (0.0007–0.0026) and that r was very high (0.9005–0.9908) for all validation, training, and test data set. The MAE value is also very low and consistent in

Fig. 2 Variations of fitness function with different generation of the GA-ANN model in training phase

118

A. Yadav et al.

Fig. 3 Performance variation with epochs in training, validation, and testing phase

Table 1 Error statistics of the proposed GA-ANN model at Waddepally gauge station in Krishna River

Statistics

Training

Validation

Testing

RMSE

0.0026

0.0007

0.0013

MAE

0.0203

0.0139

0.0182

r-correlation

0.9005

0.9908

0.9182

Error variance

0.0026

0.0007

0.0013

MSE

0.0511

0.0265

0.0364

all phases of validation, training, and testing which is varying from 0.0265 to 0.0511. It is found that, based on the RMSE, MSE, VAR, MAE, and r, the proposed GAANN model has provided more efficient and satisfactory results, with good precision for predicting the suspended sediment yield. All error statistical parameters of all three data sets have been consistent and have shown that the hybrid GA-ANN model has more generalization ability. Due to the low error values and high r values in training, validation, and testing phase, over-fitting and under-fit problems with the ANN model are avoided. To predict the suspended sediment yield at various time intervals (monthly), the optimal hybrid GA-ANN model is generated using input parameters such as water discharge and water level. The hydrograph in Fig. 4 clearly shows that the sediment yield between the observed and predicted model is closed. The scatter plot of Fig. 5 reveals that a larger number of points are even closer to the bisector axis, which concludes that the observed values of the sediment and corresponding predicted

Design of Hybrid Soft Computing Techniques for Estimation …

119

Fig. 4 Hydrograph between the observed and GA-ANN-predicted suspended sediment yield during testing phase

Fig. 5 Scatter plot between observed and GA-ANN-predicted sediment yield in training, validation, and testing phase

120

A. Yadav et al.

sediment yield by the GA-ANN model are significantly the very closed. It is also noted that all sediment values determined by the GA-ANN model are positive. In reality, sediment never be negative values. It revealed that there is more generalization potential capability in the proposed model for estimation of sediment yield. In the case of low-, medium-, and high-suspended sediment yield, a satisfactory and optimistic suspended sediment estimate was given by the GA-ANN model.

5 Conclusions The GA-ANN model uses the discharged water and water level as input factors for predicting the sediment yield at the Waddepally gauge station in the river basin of Krishna, Andhra Pradesh, India. It was observed that the most predominant parameter for estimating suspended sediment load is water discharge and water level. In order to optimize the network structure, an effective ANN structure is built in conjunction with GA to determine suspended sediment yield. From previous literature, it is observed that some soft-computing model estimated negative values during low suspended sediment but here all estimated sediment values by GA-ANN models are positive. This implies that adequate output is granted to the GA-ANN model and it has more generalization capabilities because of the simultaneous optimization of all ANN model parameters using the GA algorithm. This suggested that hybrid model can be used successfully for sediment yield calculation where sediment assessment is unavailable. The data from one gauge station and few data sets were used in the analysis. More data, including temperature, precipitation, humidity, etc., from multiple gauge stations may be needed in the future to reinforce these conclusions.

References 1. Rajaee T, Mirbagheri SA, Zounemat-Kermani M, Nourani V (2009) Daily suspended sediment concentration simulation using ANN and neuro-fuzzy models. Sci Total Environ 407(17):4916– 4927. https://doi.org/10.1016/j.scitotenv.2009.05.016 2. Zhu YM, Lu XX, Zhou Y (2007) Suspended sediment flux modeling with artificial neural network: an example of the Long chuanjiang River in the Upper Yangtze catchment, China. Geomorphology 84(1–2):111–125. https://doi.org/10.1016/j.geomorph.2006.07.010 3. Bhattacharya B, Lobbrecht AH, Solomatine DP et al (2003) Neural networks and reinforcement learning in control of water systems. J Water Res 129(6):458–465. https://doi.org/10.1061/ (ASCE)07339496(2003)129:6(458) 4. Razia S, Narasingarao MR, Bojja P (2017) Development and analysis of support vector machine techniques for early prediction of breast cancer and thyroid. J Adv Res Dyn Control Syst 9(Special Issue 6):869–878 5. Chakravorti T, Satyanarayana P (2020) Non-linear system identification using kernel based exponentially extended random vector functional link network. Appl Soft Comput 89:106–117. https://doi.org/10.1016/j.asoc.2020.106117 6. Anila M, Pradeepini G (2017) Study of prediction algorithms for selecting appropriate classifier in machine learning. J Adv Res Dyn Control Syst 9(Special Issue 18):257–268

Design of Hybrid Soft Computing Techniques for Estimation …

121

7. Yadav A, Chatterjee S, Equeenuddin SM (2017) Prediction of suspended sediment yield by artificial neural network and traditional mathematical model in Mahanadi River Basin, India. J Sustain Water Resource Manag 4(4):745–759. https://doi.org/10.1007/s40899-017-0160-1 8. Yadav A, Chatterjee S, Equeenuddin SM (2018) Suspended sediment yield estimation using genetic algorithm-based artificial intelligence models: case study of Mahanadi River, India. Hydrol Sci J 63(8):1162–1182. https://doi.org/10.1080/02626667.2018.1483581 9. Patel AK, Chatterjee S, Gorai AK (2018) Development of an expert system for iron ore classification. Arab J Geosci 11(15):401. https://doi.org/10.1007/s12517-018-3733-x 10. Ramaiah P, Kumar S (2018) Dynamic analysis of soil structure interaction (ssi) using anfis model with oba machine learning approach. Int J Civil Eng Technol 9(11):496–512 11. Patel AK, Chatterjee S, Gorai AK (2019) Development of a machine vision system using the support vector machine regression (SVR) algorithm for the online prediction of iron ore grades. Earth Sci Inf 12(2):197–210. https://doi.org/10.1007/s12145-018-0370-6 12. Patel AK, Chatterjee S, Gorai AK (2019) Effect on the performance of a support vector machinebased machine vision system with dry and wet ore sample images in classification and grade prediction. Pattern Recognit Image Anal 29(2):309–324. https://doi.org/10.1134/S10546618 19010097 13. Dabbakuti JRKK, Jacob A, Veeravalli VR, Kallakunta RK (2019) Implementation of IoT analytics ionospheric forecasting system based on machine learning and thing speak. IET Radar Sonar Navig 14(2):341–347. https://doi.org/10.1049/iet-rsn.2019.0394 14. Bisoi R, Chakravorti T, Nayak NR (2020) A hybrid Hilbert Huang transform and improved fuzzy decision tree classifier for assessment of power quality disturbances in a grid connected distributed generation system. Int J Power Energy Convers 11(1):60–81. https://doi.org/10. 1504/IJPEC.2020.104810 15. Cigizoglu HK, Kisi O (2006) Methods to improve the neural network performance in suspended sediment estimation. J Hydrol 317(3–4):221–238. https://doi.org/10.1016/j.jhydrol. 2005.05.019 16. Devadasu G, Sushama M (2016) A novel multiple fault identification with fast Fourier transform analysis. In: 1st international conference on emerging trends in engineering, technology and science, ICETETS, 2016 17. Yadav A, Satyannarayana P (2020) Multi-objective genetic algorithm optimization of artificial neural network for estimating suspended sediment yield in Mahanadi River basin, India. Int J River Basin Manag 18(2):207–215. https://doi.org/10.1080/15715124.2019.1705317 18. Yadav A, Chatterjee S, Equeenuddin SM (2020) Suspended sediment yield modeling in Mahanadi River, India by multi-objective optimization hybridizing artificial intelligence algorithms. Int J Sedim Res. https://doi.org/10.1016/j.ijsrc.2020.03.018 19. Hornik K, Stinchcombe M, White H (1989) Multilayer feed forward networks are universal approximators. Neural Netw 2:359–366. https://doi.org/10.1016/0893-6080(89)90020-8

QR-Based Ticket Verification and Parking System Mohammed Ali Hussain, K. Sree Varsha, K. Krishnamraju, K. Lavanya, and B. Chakradhar

1 Introduction Indian Railways is one amongst the organization wherever folks use the foremost to travel from completely different places [1–3]. Here, passengers want most of the protection for the persons who ought to park the vehicles there and need to travel another place through train. Indian Railway is India’s third largest human transport system over that a pair of core passengers travel daily all over India [4–6]. The passengers accomplish their journey from their supply station to destination in standing mode. The quantity of passengers in Indian Railway has been increasing drastically in per annum, throughout a rate of twenty-five to fifty from its previous year. Such increase also will increase variety of waiting queue passengers in station. Increasing number of waiting queue passengers increases rushes in train that lands up “happy journey” shibboleth of Indian Railway in to “unhappy journey” [1]. So as to ease the price, ticket booking facility for passengers, also as, boost paperless ticketing, railway is about to introduce QR code system. The national transporter is additionally operating to strengthen its Unreserved Ticketing System (UTS) for QR railway app [7]. Soon, passengers are going to be able to get QR codes of train tickets online which is foreseen to boost the paperless ticketing system. This will also be enforced within the parking stand wherever the traveller will enter into the parking stand by taking one QR code and park the vehicle there [8]. Once the passenger needs to require the vehicle out, he should scan the QR code that is given to him whereas parking the vehicle and pay the number as per the hours the vehicle is in the parking stand [9]. M. A. Hussain (B) · K. Sree Varsha · K. Krishnamraju · K. Lavanya Professor, Department of ECM, Koneru Lakshmaiah Education Foundation, Guntur, Andhra Pradesh, India e-mail: [email protected] B. Chakradhar Department of ECE, CMR College of Engineering and Technology, Kandlakoya, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_14

123

124

M. A. Hussain et al.

2 Literature Survey Mrs. Omprakash Yadav et al. check in for all the boarded passengers. If anyone willing to discontinue the journey, then check is becoming to be followed that provides the vacant seat information to the PRS and PRS will allot this seat to a non-confirmed passenger informing by an SMS [2]. In Najim sheikh et al., this bus pass ticket could also be bought with just a wise phone application, where they will carry bus pass tickets as a QR. Customer can register for a go specifying the source and destination. The appliance will generate the QR code according to the knowledge fill by users and which might be utilised by the conductor or an authorised person to scan the ticket [5]. Marika Arena et al. need developed land bus pursuit system and QR code-primarily based price ticket pass system for tracking of the buses that determines the position of the bus exploitation conductor phone via GPS and displays the position on a digital map. The foremost role performed by the Admin, Passenger, and Conductor User will search bus from supply to destination [10].

3 Related Work The existing railway price ticket generation system has several disadvantages in terms of security. Within the existing system of railways, accustomed generate tickets solely by the pc operator, the traveller fills in the queue to urge the ticket [11]. The emergency passenger cannot have that time to face inside the queue to urge the ticket before the train time of departure. Within the existing system thanks to inconvenience of a central server, the railway agents suffered unwanted delays in ticket generation and payments [12]. Within the existing system, the mixing of various train on one platform was not met [13]. With the arrival of the railway price ticket generation exploitation QR code for the unreserved class, these disadvantages are usually overcome [10, 14]. Even the QR codes that are generated at the parking stand will scale back a number of the conflicts between the vehicle house owners whereas parking the vehicle within the parking stand [15, 16].

4 Proposed Work The new projected system contains an info wherever the data of the travellers is stored [3, 9]. This database acts because the central database. This central database checks all the changes which may be done by the passenger or the user, and this could be changed or updated mechanically within the database [1, 17]. This can be the net application which can be accessed by the user terribly simply whereas booking the ticket. This methodology even reduces the work burden of the pc operator and makes

QR-Based Ticket Verification and Parking System

125

easier to the users. And even the QR code for the parking stand helps the user for the protection of the vehicles they are parking, the parking indicates days or more. This even helps within the cashless dealings too [2]. By this method, the records are often updated dynamically. This could scale back the time and cash of each traveller and additionally the pc operator. This offers security to the passenger and also the vehicles they need position if necessary. The information is simple to access, and it even reduces the labour work [18].

5 Implementation Model The whole proposed system can be divided into two modules: 1. 2.

QR code Ticket generation and scanning QR code for parking stand.

5.1 QR Code Ticket Generation and Scanning In this module, the user will book the price tag within the web application. The user enters the small print and login into the net application and book the ticket. When booking of the ticket, the user gets a QR code. This QR code is employed when the ticket checker comes in the compartment for checking the details of seats of the passenger. This makes simple for the ticket checker to envision the seats and mark them [19].

5.2 QR Code for Parking Stand In this module, the traveller will park the vehicle within the parking stand by taking the QR code whereas stepping into the stand. Once the passengers wish to require the vehicle out from the lot, the passenger have to be compelled to scan the QR code. Once the person scans the QR code, the person will be able to check the fare for the vehicle to be position in the parking stand [20] (Fig. 1).

126

M. A. Hussain et al.

Fig. 1 QR code

TICKET COLLECTOR QR VERIFIER INVALID FINE

SCAN QR FROM PASSENGER VALID VERIFIED Fig. 2 QR code verification

6 Block Diagram for Proposed Model 6.1 QR Code Verification As per Fig. 2, it shows that porter encompasses a device to scan the QR code having with the traveller. If the QR code is with success verified, then it is marked else the passenger must pay the penalty because the person is not having the price tag of that individual seat.

6.2 Flow Chart for Parking Allotment and Payment As per Fig. 3, it shows that when the client enters the railway station client checks whether he want to go for parking or buy a ticket. If he chooses for the parking he

QR-Based Ticket Verification and Parking System

127 NO GO TO THE COUNTER

START NO

PARKING

TICKET

YES NO

GO TO YYESPLATFORM

CHECK FOR AVAILABILITY

YES PARK IN THE SLOT

Fig. 3 Parking allotment

checks for the availability if yes, he parks the vehicle and goes for the ticket else directly go for the ticket. If the person has the ticket, i.e. QR code, then the person goes to platform else the person goes to the ticket counter to purchase the ticket [22].

6.3 Pseudo Code for Ticket Verification Step 1: The Ticket collector will have a device to verify the QR code ticket. Step 2: Once, the Ticket collector approaches the passengers for ticket verification. The ticket which has QR code will be scanned by the ticket collector. Step 3: If QR is verified successfully then it is marked else the ticket collector asks for penalty. Step 4: End.

6.4 Pseudo Code for Parking System Step 1: Check for the parking availability and go to that place and scan the QR code in that slot. Step 2: Park your vehicle after scanning the QR code.

128

M. A. Hussain et al.

Step 3: The timer will start when the QR code is scanned. Step 4: Once, we want to get out from the parking we will again scan the QR code then we will get the fare to pay. The pay can be calculated as: If no. of hours > 10 then bill = bill + (no. of hours) *10 Else if no. of hours < 10 then bill = bill + (no. of hours) *5 Else id no. of hours > 24 then bill = bill + (no. of hours) *20 And so on. Step 5: Pay the bill using the QR code that are kept like Phonepe, Google Pay, etc. Step 6: End.

7 Experimental Results As per Fig. 4, it is one of the results where we have acquired from the given problem statement. This result shows how the data is read from the QR code which has been generated before. As per the given Fig. 5, it shows that the generation of QR code is successfully done. This QR code is generated to the data which is given or for the links which are generated through the data.

Fig. 4 Reading a QR code

Fig. 5 QR created from the links

QR-Based Ticket Verification and Parking System

129

8 Conclusion In this paper, we have developed Internet application exploitation MYSQL and ZXING library for QR code generation, and Java for the server aspect that is that the approach; however, the price tickets are booked. This kind of reservation and conjointly allotment for the parking are often employed in any variety of transport system. This application can save the time of each user and also the ticket counter members. For the validation of the QR code, we are able to use hardware devices within the lot and also with the ticket checker. This may be very ease for the user to require trains as this application can facilitate the user to induce the closest location. This application is extremely compatible to use. Hence, this may solve a number of the problems that were there within the previous approach.

References 1. Deshpande PR, Mujumdar TN, Sarode SR, More SB (2018) Online ticket substantiation using QR code based android application system. Int Res J Eng Technol (IRJET) 2. Yadav O, Fernandes R, Tiwari R, Kaul S (2014) Online reservation system using QR code based android application system. Int J Sci Res Publ 3. Gunjan VK, Pathak R, Singh O (2019) Understanding image classification using TensorFlow deep learning-convolution neural network. Int J Hyperconnectivity Internet Things (IJHIoT) 3(2):19–37 4. Gao R, Zhao M, Ye T, Ye F, Wang Y, Luo G (2017) Smartphone-based real time vehicle tracking in indoor parking structures. IEEE Trans Mob Comput 16(7):2023–2036 5. Sheikh N, Shende M, Pandit R, Samart S, Khapekar T, Kumar S, Kumar V (2018) QR based E-ticket system. Int J Future Revolution Comput Sci Commun Eng 6. Waghmare A, Pansambal S, Pavate A, Kumawat D (2018) QR code based Railway e-Ticket. IOSR J Eng (IOSR JEN) 7. Rajesh K, Waranalatha SS, Mounlka Reddy KV, Supraja M (2018) QR code-based real time vehicle tracking in indoor parking structures. IEEE international conference 8. Ahmed SM, Kovela B, Gunjan VK (2021) Solar-powered smart agriculture and irrigation monitoring/control system over cloud—an efficient and eco-friendly method for effective crop production by farmers in rural India. In: Proceedings of international conference on recent trends in machine learning, IoT, smart cities and applications. Springer, Singapore, pp 279–290 9. Rouillard J (2014) Contextual QR codes. IEEE Xplore 10. Ashitha KS, Rafeeda K (2017) Message sharing and document authentication using QR code. Int J Inf Sci Appl 11. Tiwari S (2016) An introduction to QR code technology. In: International conference on information technology (ICIT) 12. Balakrishna S, Solanki VK, Gunjan VK, Thirumaran M (2019) Performance analysis of linked stream big data processing mechanisms for unifying IoT smart data. In: International conference on intelligent computing and communication technologies. Springer, Singapore, pp 680–688 13. Arena M, Foiadelli F, Acquaro G, Gentile M (2015) Functional safety of railway systems. AEIT international annual conference (AEIT) 14. Bani-Hani RM, Wahsheh YA, Al-Sarhan MB (2014) Secure QR code system. In: 10th international conference on innovations in information technology (IIT) 15. Michael D’silva G, Kunjumon Scariah A, Roy Pannapara L, John Joseph J (2017) Smart ticketing system for railways in smart cities using software as a service architecture. International conference on I-SMAC (IoT in social, mobile, analytics and cloud) (I-SMAC)

130

M. A. Hussain et al.

16. Devadasu G, Sushama M (2016) A novel multiple fault identification with fast Fourier transform analysis. In: 1st international conference on emerging trends in engineering, technology and science, ICETETS 17. Tharani Sri Sakthi BT, Leo JJ, Monisha R, Ramesh SM (2014) Advanced train reservation and passenger intimation with safety system. In: International conference on information communication and embedded systems (ICICES2014) 18. Khanna A, Anand R (2016) IoT based smart parking system. In: International conference on internet of things and applications (IOTA) 19. Ma S, Jiang H, Han M, Xie J, Li C (2017) Research on automatic parking systems based on parking scene recognition. IEEE Access (volume 5)

Design and Analysis of Uniformly Illuminated 8-Way Wilkinson Power Divider for L-Band Applications K. T. P. S. Kumar, Lakshman Pappula, Vankayalapati Sahiti, M. V. Sai Kalyan, Nallamalapu Pratapreddy, and Potharaju Vinay Kumar

1 Introduction The communication systems are part and parcel of diverse applications. One such field is microwave applications. The microwave applications include the usage of power dividers in the construction of the antenna arrays and power amplifiers. In specific, the power dividers are used at feeding networks of the antenna arrays, also in phase shifters setup. The power divider is applied wherever the exploitation of channels is concerned. In wireless communication, the power divider is used whenever there is a need to divide the power either equally or unequally into smaller amounts. Among the many kinds of power dividers, the concentration of the current work is on Wilkinson power dividers as they have the advantage of uncomplicated geometry, and the size is compact and negligible insertion loss on par with others. The property of this divider in providing the better isolation in-between the output ports while maintaining matched impedance on all the ports has attracted the author to work on this design. Wilkinson power divider can be used either at the input/output port for splitting or recombination. The perfect divider characteristics [1–4] are obtained by considering the improvement of transmission line segment properties. It consists of basic level components like transmission lines and resistors. Even though this method does not extend to flat dividers, it is well appropriate for impedance matching across all the available frequency bands [5]. A further explanation that separates such power divisors is that separation is the most critical factors in microwave circuits, as all ports in microwave circuits are made possible by such dividers [6, 7]. On the other hand, microstrip power divider circuits with high power division rate are needed in many applications [8]. The structure of a Wilkinson power mainly consists of three ports formed as a network and the output ports are matched losslessly, and dissipation occurs due to reflected power. Output control of phase signals of the K. T. P. S. Kumar · L. Pappula · V. Sahiti (B) · M. V. Sai Kalyan · N. Pratapreddy · P. V. Kumar Koneru Lakshmaiah Education Foundation, Vaddeswaram, Andhra Pradesh 522302, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_15

131

132

K. T. P. S. Kumar et al.

Fig. 1 Structure of a Wilkinson power divider

constant amplitude may be broken into two or more. A π /4 impedance transformer is utilized for a classic impedance of 2Z0 when the construction is of a two-way power divider. Basically, all the Wilkinson Power Dividers are known as microstrip lines in the field of microwave engineering as seen in figure. Figure 1 [9, 10] clearly outlines the impedance requirement of input transmission line. The factor which is placed over the transmission line is multiplied with the characteristic of input transmission line. This satisfies the impedance requirement of every one fourth part of wave transmission. Another point of concentration here is that the output ports are linked by an input resistor whose transmission line impedance at the input end is compounded by a factor two. Such impedances require an isolation and balancing of the outputs of the Wilkinson power divider [11, 12]. The note to be made here is that whenever the signal input is given at port two, the signal that is incident at port one is only half of the total strength. The reciprocity (S21 = S12) holds good [13] though half of the original power is dissipated into the capacitor. The condition of setting S23 and S32 at zero has the following phenomenon, for Wilkinson power divider optimal input is applied at port 2 as the result no power is detected at port three [14]. This is due to the fact that the ports, port2 and port3 are separated by a distance. Also, the Wilkinson power divider is able to manage reflections very well inside the device [15, 16]. Here, we have the exact replica of the signal being placed at the output port that may also be told as reflection equivalent of the actual signal located at the output port [17]. The mirrored signal passes along two pathways back to the other output port: The resistor and the input junction are the two pathways. As the transmission lines are of one fourth-wavelength each, the signal enters a particular phase via the resistor, while the pulse moving down the two quarter-wave lines (two phase changes of 90°) crosses out of phase 180° [18, 19]. If such reflections are of similar amplitude out of phase, they result in total cancelation (the separation between the ports of the output). The Wilkinson power divider offers isolation in electrical form but not any physical isolation [20]. This is eventually achieved by creating an equal power splitter that retains binary topology and adjusts the width of the microstrip lines to represent the equal power splitter and estimate the cstschema. Wilkinson’s equivalent power splitter is designed by using RT Duroid, and the geometric measurements are as follows: thickness of the substrate is taken as 1575 mm, the dielectric constant value is given by εr = 2.33, and at the feed part, the thickness of copper is given by 0.035.

Design and Analysis of Uniformly Illuminated 8-Way …

133

2 The Wilkinson Power Divider with Numerical Illustrations See Table 1.

2.1 Effective Dielectric Constant On the off chance that we increment the width of the microstrip, an ever-increasing number of electric lines of powers are inside the dielectric field. The dielectric constant along these lines increments with the expansion in the fellow width. If there is an occurrence of the microstrip, some part of the electric field is in air and a section is in the dielectric. If the dielectric steady of the material is 1r , the dielectric constant of the structure for the computation of capacitance per unit length is somewhere close to 1 and 1r . The following formula [5] can be used to determine the effective dielectric constant ⎞

⎛ εeff =

εr + 1 εr − 1 ⎝ 1 ⎠ + ∗ / 2 2 1+12D W

where W D 1r

represents width of the microstrip represents thickness of dielectric represents relative permittivity

2.2 Characteristic Impedance Characteristic impedance (z0 ) is the ratio of the amplitude of voltage to the current of single wave propagating along the line that is a wave traveling in one direction. The formula for characteristic impedance is Table 1 Material dimensions

S. No.

Parameter

Value

1

Length of 50 ohms strip

29.75 mm

2

Width of 50 ohms strip

4.67 mm

3

Length of 70 ohms strip

32.29 mm

4

Width of 70 ohms strip

2.3 mm

5

Input impedance

50 Ω

134

K. T. P. S. Kumar et al.

Z 0 = Vi (t)/Ii (t)  Z 0 = Z 0o Z 0E Regardless of the length, the characteristic impedance is constant. It is expressed in ohms but cannot be measured by ohmmeter. Here, characteristic impedance is 50 Ω.

2.3 Width of the Microstrip W = D



W 8e A A −2 for D e2 2 B−1− π

2



where W D

represents conductor width represents Dielectric substrates thickness

  Z o εr + 1 εr − 1 0.11 A= + 0.23 + 60 2 εr + 1 εr and 377π B= √ 2Z 0 εr

where Z0

is taken as impedance which is 50 Ω

2.4 Length of the Microstrip Length of the strip is considered as 1/4th of the wavelength, and this wavelength can be obtained from the formula λm =

300 √ F εeff

where λm F

represents wavelength represents the desired frequency, i.e., 1.8 GHz

Design and Analysis of Uniformly Illuminated 8-Way …

135

The length and width are separately calculated for both 50 ohms and 70 ohms strips. The 2-way Wilkinson Power Divider (Fig. 2): The results are depicted at the below figures as shown (Figs. 3, 4, and 5):

Fig. 2 2-way Wilkinson power divider

Fig. 3 Reflection coefficient S11 at input port

Fig. 4 Insertion loss at output port (S21 )

136

K. T. P. S. Kumar et al.

Fig. 5 Isolation loss at output ports of power divider (S2,3 )

3 The Proposed Design and Analysis The design procedure of an 8-way Wilkinson power divider requires setting up the parameters of the microstrip. This is done by denoting the length and the width which follows the formulae mentioned in section II. A distance of lambda/2 is maintained between two successive power dividers in the top, and as we move downwards, the length of the 70 Ω strip gets doubled and the same λ/2 distance between the two successive power dividers is maintained. The horizontal microstrips are of impendence 70.7 Ω, whereas the vertical ones are of 50 Ω. The width and length of different strips are calculated as below (Fig. 6; Table 2).

Fig. 6 8-way Wilkinson power divider

Table 2 Design specifications

S.No.

Structural element

Value (mm)

1

Width of 50 ohm strip (W )

4.67

2

Length of 50 ohm strip (L)

29.75

3

Width of 70 ohm strip (W 1)

2.3

4

Length of 70 ohm strip (L1)

32.29

5

Length of 70 ohm strip (L2)

64.59

6

Length of 70 ohm strip (L3)

129.16

Design and Analysis of Uniformly Illuminated 8-Way …

137

4 Results and Discussions 4.1 S-Parameters S11 represents the total amount of power the antenna reflects and thus is known as the coefficient of reflection (sometimes written as gamma or loss of return). If S11 = 0 dB, then the antenna represents all the power, and nothing radiates. The other condition where the S11 = 10 dB is obtained the result interpretation is as below 7 dB of the power is reflected away while the 3 dB of it is delivered to the antenna. The majority of the power had been recognized by or delivered to the antenna (Fig. 7). At the operating frequency, we can observe the reflection coefficient is below − 10 dB which indicates the wave is transmitted without much reflection. Here, the S11 parameter is noted as −20 dB [21]. S21 denotes the power transmitted from port1 to port2. In other words, it may be described as forward voltage gain transmitted. As a result, it is a proportion of signal coming out from port2 comparative through (by) the sign entering port1. Likewise, the other parameters s31 to s91 are also the transmission coefficients with respect to port1 [22] (Fig. 8).

Fig. 7 Reflection coefficient s11 at Input Port

Fig. 8 Transmission coefficient at output end

138

K. T. P. S. Kumar et al.

Fig. 9 E-Field with a maximum value of 4192.63 V/m

For a 2-way power divider, the transmission coefficient must be below −3 dB whereas for an 8-way power divider it must be below −9 dB. Here, all the transmission coefficients from S21 to S91 are below −9 dB at the desired frequency.

4.2 Electric Field (E-field) In a far field region, the electric filed intensity can be estimated by considering the force which is being applied by it on a unit point charge. The electric field intensity can also be estimated by voltage as measure. If there exists a large voltage difference between terminals of an antenna, then we can draw a conclusion that there exists a strong electric field (Fig. 9).

4.3 Magnetic Field (H-Field) The propagation of the waves in the far field region is the result of the E-field and H-field, and the H-field is orthogonal to the plane of propagation of a plane wave and is perpendicular with respect to the electric field (Fig. 10).

Fig. 10 H-field maximum magnetic field at 34.617 V/m

Design and Analysis of Uniformly Illuminated 8-Way …

139

Fig. 11 Surface current with a maximum value of 16.043 A/m

4.4 Surface Current In antennas, the surface current is a genuine electric current that is actuated by an applied electromagnetic field. To make radiation, there must be charge increasing speed and de-acceleration and these charges establishes a current. So on a strip how the charges get gathered and where they are collecting is clarified by the surface current (Fig. 11).

5 Conclusion In this paper, an 8-way Wilkinson power divider design and its analysis is shown using Computer Simulation Technology (CST) software. This kind of power divider is suitable to be united with array feeding network. Each part in this prototype is separately tested and designed in order to obtain equal power dividing at each output port. There will be small deviations from the theoretical calculations which are modified by using samples in CST. Thus, the designed 8-way Wilkinson power divider depicts good quality matching and isolation with almost precisely dividing ratios at all the output ports.

References 1. Pappula L, Ghosh D (2016) Synthesis of thinned planar antenna array using multiobjective normal mutated binary cat swarm optimization. Appl Comput Intell Soft Comput 2016:1–9. https://doi.org/10.1155/2016/4102156 2. Hallberg W, Ozen M, Kuylenstierna D, Buisman K, Fager C (2018) A Generalized 3-dB Wilkinson power divider/combiner with complex terminations. IEEE Trans Microw Theory Tech 66(10):4497–4506. https://doi.org/10.1109/TMTT.2018.2859305 3. Kshitija T, Ramakrishna S, Shirol SB, Kumar P (2019) Micro-strip patch antenna using various types of feeding techniques: an implementation. Proc Int Conf Intell Sustain Syst ICISS 2019, pp 318–322. https://doi.org/10.1109/ISS1.2019.8908066

140

K. T. P. S. Kumar et al.

4. Chen A, Zhuang Y, Zhou J, Huang Y, Xing L (2019) Design of a broadband Wilkinson power divider with wide range tunable bandwidths by adding a pair of capacitors. IEEE Trans Circuits Syst II Express Briefs 66(4):567–571.https://doi.org/10.1109/TCSII.2018.2803076 5. Shaikh FA, Khan S, Alam AHMZ, Habaebi MH (2019) Design an ultra-wideband modified Wilkinson power divider fed-by balanced Antipodal Vivaldi antenna array for imaging applications. Int J Eng Adv Technol 9(2):4236–4241. https://doi.org/10.35940/ijeat.b4936. 129219 6. Oraizi H, Sharifi AR (2006) Design and optimization of broadband asymmetrical multisection Wilkinson power divider. IEEE Trans Microw Theory Tech 54(5):2220–2231. https://doi.org/ 10.1109/TMTT.2006.872786 7. Pozar DM (2011) Microwave engineering (4th edn). Printed in the United States of America: ISBN: 978-0-470-63155-3 8. Heydari M, Roshani S (2017) Miniaturised unequal Wilkinson power divider using lumped component elements. Electron Lett 53(16):1117–1119. https://doi.org/10.1049/el.2017.2118 9. Siragam S, Dubey RS, Pappula L (2019) Synthesis and study of zinc titanium aluminate nanoceramic composite for patch antenna application. In: 2019 5th IEEE international WIE conference on electrical and computer engineering, WIECON-ECE 2019—proceedings, pp 47–50. https:// doi.org/10.1109/WIECON-ECE48653.2019.9019915 10. Pappula L, Ghosh D (2018) Cat swarm optimization with normal mutation for fast convergence of multimodal functions. Appl Soft Comput J 66:473–491. https://doi.org/10.1016/j.asoc.2018. 02.012 11. Kurniadi DP (2012) Design and realization Wilkinson power divider at frequency 2400 MHz for radar S-band. IOSR J Electron Commun Eng 3(6):26–30. https://doi.org/10.9790/2834-036 2630 12. Shabankareh MAG, Arman E (2019) Simulation and fabrication of a Wilkinson 8-port unequal power splitter. Univers J Electr Electron Eng 6(1):23–30. https://doi.org/10.13189/UJEEE. 2019.060103 13. Qaroot AM, Dib NI (2010) General design of N-way multi-frequency unequal split Wilkinson power divider using transmission line transformers. Prog Electromagn Res C 14:115–129. https://doi.org/10.2528/PIERC10060109 14. Ahmed SM, Kovela B, Gunjan VK (2020) IoT based automatic plant watering system through soil moisture sensing—a technique to support. Adv Cybern Cogn Machine Learn Commun Technol 643:259 15. Bao C, Wang X, Ma Z, Chen CP, Lu G (2020) An optimization algorithm in ultrawide band bandpass Wilkinson power divider for controllable equal-ripple level. IEEE Microw Wirel Components Lett 30(9):861–864. https://doi.org/10.1109/LMWC.2020.3011516 16. Lakrit S, Medkour H, Das S, Madhav BTP, Ali WAE, Dwivedi RP (2020) Design and analysis of integrated Wilkinson power divider-fed conformal high-gain UWB array antenna with band rejection characteristics for WLAN applications. J Circuits Syst Comput. https://doi.org/10. 1142/S0218126621501334 17. Wu Y, Jiao L, Zhuang Z, Liu Y (2017) The art of power dividing: a review for state-of-the-art planar power dividers China. Communication 14(5):1–16. https://doi.org/10.1109/CC.2017. 7942190 18. Kalpanadevi MGE, Nishaw MKN, Priyamalli E, Radhika V, Shenbaga Priyanga V (2017) Design and analysis of Wilkinson power divider using microstrip line and coupled line techniques, vol 34, p 2017. [Online]. Available: www.iosrjournals.org 19. Dharma Raj C, Sasibhushana Rao G, Jayasree PVY, Srinu B, Lakshman P (2010) Development of a three layer laminate for better electromagnetic compatibility performance at X-band. Commun Comput Inf Sci 101:406–410. https://doi.org/10.1007/978-3-642-15766-0_64 20. Li M, Luk KM (2014) Low-cost wideband microstrip antenna array for 60-GHz applications. IEEE Trans Antennas Propag 62(6):3012–3018. https://doi.org/10.1109/TAP.2014.2311994 21. Yuwono R, Ramadhan W, Asmungi G (2016) Design and prototype of microstrip power divider for analog and digital television antenna applications at the frequency of 479–799 MHz. ARPN J Eng Appl Sci 11(1):712–715

Comparative Analysis on Mulberry Leaf Disease Detection Using SVM and PNN Y. Rakesh Kumar, P. Satyanarayana Goud, and Sheelam Pravalika

1 Introduction Agricultural productivity is one of the main contributes to India economy. So detection of plant diseases is a significant task in agriculture field. Automatic disease detection at initial stage is beneficial. Manual plant disease detection is more laborious task, is less accurate, requires a large team of experts, and costs very high for large farms. An automatic detection technique is acquired which requires less time, less manual intervention, and accuracy is good. In plants, generally fungal, bacterial, viral diseases, yellow and brown spots are observed. Image segmentation is used to separate the interest part in different application. There are different segmentation methods which include ROI, clustering thresholding, and advanced deep learning methods. Segmentation is done based on the features like texture, color, shape, and boundary, which are extracted from interested diseased part. An automatic image segmentation and classification is a main step in computeraided image processing. This work aims to compare SVM and probabilistic neural networks (NNs) in detecting plant diseases in the classifier stage.

2 Literature Survey Two feature extraction techniques are presented in [1]. First, 12 texture features are estimated by using Gray Level Covariance Matrix (GLCM) for diagnosis purpose. In second approach, AlexNet, a pretrained deep learning model, is used to extract Y. Rakesh Kumar (B) · P. Satyanarayana Goud · S. Pravalika ECE, G. Narayanamma Institute of Technology and Science (for Women), Hyderabad, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_16

141

142

Y. Rakesh Kumar et al.

1000 features automatically. Here, back propagation neural network (BPNN) is used as classifier and observed 93.85% accuracy by deep learning features compared to the texture features. Al-Hiary et al. [2] K-means clustering [3, 4] is used for segmentation with masked green pixels and Otsu threshold method [5] then neural network with SGDM texture analysis is used for classification. Ranjan et al. [6] Segmentation is done based on HSV color features, and then artificial neural network is used to classify healthy and unhealthy samples. Similarly, Kulkarni et al. [7] used ANN with color features extracted from Gabor filter [8]. Khirade et al. [9] presented various segmentation methods to extract disease part of the plant leafs. K-means clustering followed by Ostu algorithm is used to segment disease part. GLCM features are estimated for classification. ANN method is used after features are extracted for classification of diseases in plants, i.e., modified SOM, back propagation, and multiclass support vector machines are used. Sannakki et al. [10] used the Feed Forward Back Propagation Neural Network technique to diagnose and classify diseases in grapes. The author used the pictures of the grape leaf with a complex background for diagnosis. Downy mildew and powdery mildew images are considered to check the performance. Other anisotropic streams are used to suppress noise and then segmented using K-means and classified using neural network. To validate the results, confusion matrix and accuracy are considered. Al Bashish et al. [11] also used K-means and classified using neural network for leaf and stem disease detection. Leaf diseases considered are early scorch, ashen mold, late scorch, cottony mold, and tiny whiteness images for testing. Mustafa et al. [12] used neural network classification and compare the feed forward neural network and radial basis function neural network analyzing on citrus trees’ disease conditions in outdoor environment. Kutty et al. [13] used the neural network to categorize diseases in late blight watermelon leafs and anthracnose. Infected leafs are captured under controlled environment. Region of interest is utilized for color feature extraction, and Neural Network Pattern Recognition and Statistical Package for Social Science Toolbox of MATLAB is used for disease classification. Based on RGB mean color component, 75.9% accuracy is achieved. Yao et al. [14] presented detecting diseases in rice crop using color transform and Otsu algorithm and SVM. Spot disease on rice crop is segmented and extracted their texture and shape features. Both color and shape features are selected for classification because light influences color features. The SVM [14–16]Technique is used to categorize rice sheath blight and rice blast, bacterial leaf blight. Garcia et al. [16] presented a plant disease detection, quantify and classification a review. Digital images in invisible spectrum are assumed. Gill [17] presented k mean clustering for leaf segmenting and removing background. Then, Gray Level Co-occurrence Matrix (GLCM) Haralick texture features are used to classify diseased area.

Comparative Analysis on Mulberry Leaf Disease Detection …

143

3 Plant Disease Identification There are several methods existing for the segmentation of plant disease detection. The existing method, support vector machine (SVM) is explained and compared with probabilistic neural network [16, 12]. Support Vector Machine (SVM)-based leaf disease classification One of supervised machine learning algorithm is support vector machine used for both regression and classification. But most use of it is in classification tasks like texture classification and pattern recognition tasks. Data is mapped nonlinearly to data which is linearly separated in high dimensional space. Marginal distance between different classes is maximized [16]. A multiclass SVM one to rest approach is used, in this hyper plane which separates one class from all other at once. Here, separation takes all point into account and divides them into two groups. The SVM workflow is presented as: Step 1: Read input image. Step 2: Input image is mapped into HSV (Hue Saturation Value) image. Step 3: Diseased part segmentation can be achieved by the help of K-means Clustering method. Step 4: Gray Level Co-occurrence Matrix (GLCM) features are calculated for segmented image. Step 5: SVM with GLCM Features is applied to classify the leaf disease. Probabilistic Neural Network (PNN) PNN is used for pattern recognition and classification tasks. Parent probability distribution function in PNN of each class is represented by anon–parametric function and Parzen window. The class probability of a new data is calculated using PDF of each class and to allocate Bayes rule is employed to allocate the class with highest probability. It is derived from Kernel Fisher and Bayesian network. PNN has four layers of nodes: • • • •

Input layer Hidden layer Summing layer Output layer

Input layer—It is N input feature vector layer; these input neurons feed the neurons of hidden layers. Hidden layer—Neuron of these layer estimates the test samples. Euclidean distance from neurons’ center point and then radial basis function using sigma values is applied. Each hidden layer group leads to kth classier N neurons which lead to K classes. Hidden layer estimates Gaussian for each feature vector.

144

Y. Rakesh Kumar et al.

If s exemplar Feature vectors [x (s) ; s = 1, … S] labeled as class 1 and t exemplar feature vectors [y(t) ; t = 1, …, T ] labeled as class 2, likewise class k can be defined. The Gaussian centered on class 1 and class 2 for feature vectors x (s) and y(t) can be equated as in Eqs. (1), (2), respectively.      √ 2 G1(x) = 1/ (2π σ 2 ) N exp −x − x (s)  /(2σ 2 ) (1)   √    2 G2(y) = 1/ (2π σ 2 ) N exp − y − y (t)  /(2σ 2 ) (2)      y − y (s) 2 f 1 (x) = [1/(2π σ ) ](1/S) exp − 2σ 2 (s−1,S) 2 N



     y − y (t) 2 f 1 (y) = [1/(2π σ ) ](1/T ) exp − 2σ 2 (t−1,T ) 2 N



(3)

(4)

Output layer—It sums all neurons of hidden layers and sum forms a PDF Probability Density Function. The parent windows or mixed Gaussians sums are defined Eqs. (3), (4), respectively.

4 Results and Discussion To check the proposed method’s performance, experiments were carried out on mulberry diseased leaf images of leaf spot, powdery mildew, and root rot. Diseased leaf images collected were around 150 and 50 of non-diseased, from which 150 were randomly selected for training purpose and remaining were used to test the performance of SVM and PNN. Firstly, images are resized to 128 × 128 and then converted to single band image. To decrease the computation time, mulberry diseased part is segmented from diseased image of mulberry leafs. By employing K-means clustering method, diseased part is segmented for SVM classifier and statistical mean local thresholding is employed to segment diseased part for PNN classifier. To classify the disease, GLCM features such as Energy, Contrast, Entropy, Correlation Coefficient, and Homogeneity were estimated. Comparative result analysis on diseased and non-diseased leaf of mulberry tree for both SVM and PNN classifier is summarized in Table 1. It is observed for 200 diseased and non-diseased leafs SVM correctly detects around an average of 47 out of 50 images, which shows an average accuracy of 93.5% calculated based on (5).

Comparative Analysis on Mulberry Leaf Disease Detection …

145

Table 1 Comparative results of SVM and PNN Diseased leaf

Total leaf

SVM detected correctly

SVM accuracy (%)

PNN detected correctly

PNN accuracy (%)

Non-diseased

50

46

92

49

98

Leaf spot

50

47

94

49

98

Powdery mildew

50

46

92

48

96

Root rot

50

48

96

49

98

Fig. 1 a Original mulberry diseased leaf. b Gray image. c Statistical mean threshold segmented image. d Disease localized image

PNN correctly detects around an average of 49 out of 50 images, which shows an average accuracy of 98%. Accuracy can be improved by increasing image dataset. Figure 1 shows the stage wise result and images for PNN classifier. Accuracy =

Number of particular diseased image detected correctly Total number of particular diseased images

(1)

5 Conclusion The proposed neural network is used as a classifier to identify the diseases of plants if affected. The proposed method gives us accurate results and less complex compared to SVM. As it is sufficient to train the NN classifier only once, it has the ability to learn the tasks for the data given during training or learning process if the database is increased. Average accuracy of SVM is 93.5% whereas for PNN is 98% for 200 samples.

146

Y. Rakesh Kumar et al.

References 1. Sapkal AT, Kulkarni UV (2018) Comparative study of leaf disease diagnosis system using texture features and deep learning features. Int J Appl Eng Res 13(19):14334–14340 2. Al-Hiary H, Bani-Ahmad S, Reyalat M, Braik M, Alrahamneh Z (2010) Fast and accurate detection and classification of plant diseases. IEEE-IJCA 17(1):31–38 3. Mainkar PM, Ghorpade S, Adawadkar M (2015) Plant leaf disease detection and classification using image processing techniques. Int J Innov Emerg Res Eng 2(4):139–144 4. Amoda N, Jadhav B, Naikwadi S (2014) Detection and classification of plant diseases by image processing. Int J Innov Sci Eng Technol 1(2) 5. Naikwadi S, Amoda N (2013) Advances in image processing for detection of plant diseases. Int J Appl Innov Eng Manag 2(11):168–175 6. Ranjan M, Weginwar MR, Joshi N, Ingole AB (2015) Detection and classification of leaf disease using artificial neural network. Int J Tech Res Appl 3(3):331–333 7. Kulkarni AH, Ashwin Patil RK (2012) Applying image processing technique to detect Plant disease. Int J Mod Eng Res 2(5):3661–3664 8. Narayana VA, Premchand P, Govardhan A (2009) A novel and efficient approach for near duplicate page detection in web crawling. In: 2009 IEEE international advance computing conference, IACC 2009 9. Khirade SD, Patil AB (2015) Plant disease detection using image processing. In: International conference on computing communication control and automation 2015. IEEE, pp 768–771 10. Sannakki SS, Rajpurohit VS, Nargund VB, Kulkarni P (2013) Diagnosis and classification of grape leaf diseases using neural networks. In: Computing communications and networking technologies (ICCCNT). IEEE, pp 1–5 11. Al Bashish D, Braik M, Ahmad SB (2010) A frame-work for detection and classification of plant leaf and stem diseases. In: IEEE international conference on signal and image processing, pp 978–982 12. Mustafa Choudhary G, Gulati V (2015) Advance in image processing for detection of plant diseases. Int J Adv Res Comput Sci Softw Eng 5(7):1090–1093 13. Kutty SB, Abdullah NE, Hashim H, Sulinda A (2013) Classification of watermelon leaf diseases using neural network analysis. In: 80 business engineering and industrial applications colloquium 2013. IEEE, pp 459–464 14. Yao Q, Guan Z, Zhou Y, Tang J, Hu Y, Yang B (2009) Application of support vector machine for detecting rice diseases using shape and color texture features. In: International conference on engineering computation 2009. IEEE computer society, pp 79–83 15. Arivazhagan S, Newlin Shebiah R, Ananthi S, Vishnu Varthini S (2013) Detection of unhealthy region of plant leaves and classification of plant leaf diseases using texture features. Agric Eng Int CIGR 15(1):211–217 16. Garcia J, Barbedo A (2013) Digital image processing techniques for detecting, quantifying and classifying plant diseases. Springerplus 2: 660:1–12 17. Gill GS (2020) Detection of diseased section in leaves using image processing. Int J Adv Res Eng Technol 11(7):296–305 18. Ghaiwat SN, Arora P (2014) Detection and classification of plant leaf diseases using image processing techniques: a review. Int J Recent Adv Eng Technol 2(3):2347–2812 19. Ahmed SM, Shaik F, Gunjan VK, Ali MY (2021) A literature survey on identification of asthma using different classifier and clustering techniques. Mod Approaches Mach Learn Cogn Sci A Walkthrough 69–80 20. Prasad PS, Gunjan VK, Pathak R, Mukherjee S (2021) Applications of artificial intelligence in biomedical engineering. In: Handbook of artificial intelligence in biomedical engineering. Apple Academic Press, pp 125–145

Thermal Analysis of Different Components on the PCB Using ANSYS Software K. Sanjitha, V. Panduranga, and S. Mallesh

1 Introduction One of the major challenges faced in advanced electronic systems is how to reduce the temperature of an integrated circuit for high electrical–mechanical reliability, because as the integrated circuit size becomes slim and small, the number of transistors accommodated per unit area increases and thus high transfer speed of information takes place and thus the temperature increases [1, 2]. Power dissipation performance should be understood before integration devices on a printed circuit board (PCB) to confirm that all devices are operated within its outlined temperature limits. Moreover, resistance of a wire will increase as temperature will increase that successively mean that the delay increases as temperature rises [3]. Temperature has an adverse impact on the reliability of devices, with the raise in temperature metal migration increases and causes the metal wires to become thinner, which is nothing but the self-inductance of these metal wires will increase and gives rise to increase in magnetic coupling, which causes the early ageing of the devices [4, 5]. The temperature of a block depends on the activity profile of the block and the temperature of its neighbouring blocks [6]. Hence, floor-planning and placement of electronic elements have a robust impact on the thermal profile of the assorted blocks in the PCB [7]. On a macro-scale, engineers are looking at predicting and resolving the thermal hotspots in the design stages as early as possible [8, 9]. Although the idea is not new, semiconductor companies are researching for a much more effective methodical approach for estimating thermal profile and hot spots through emulation and

K. Sanjitha · V. Panduranga (B) CMR College of Engineering and Technology, Hyderabad, Telangana, India e-mail: [email protected] S. Mallesh CMR Technical Campus, Hyderabad, Telangana, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_17

147

148

K. Sanjitha et al.

simulation vectors [10, 11]. These can be used to perform a thermally aware placement, and this concept can even be extended to perform software optimizations to minimise thermal issues [12].

2 Thermal Analysis of PCBs Power dissipation plays a major role in the advanced PCB designs. Power dissipation will result in temperature differences in the chip and will reflect a thermal problem to a chip like reliability issues and also due to the heat dissipation the electrical performance also is affected and in some cases the device electrical safety is also not assured [13]. Hence, the overall temperature of the chip should be kept almost constant and must be less than the maximum ambient allowable. The electronic products’ thermal integrity mostly depends on the Printed Circuit Board which holds the electronic components like ICs, FPGAs or ASIC. Furthermore, as the size of the chip is decreasing, the heat generated is getting concentrated within a smaller region which gives rise to increase in power density at a particular region. Since the chip size is shrinking, the transistors’ size also decreases and is densely packed in a chip and the higher switching speed of the clock is also the reason for the power dissipation [14]. The traditional thermal heat transfer theory is applied for the thermal analysis of the PCB or full-chip analysis [15]. The law of heat conduction defined by Fourier describes the classic thermal heat transfer in a PCB or a chip; according to law, the heat flux vector, ‘q’ which has the units Watts per metre square, is directly proportional to the temperature’s negative gradient ‘∇T ’ expressed in Kelvin, and the proportionality constant is the materials’ thermal conductivity ‘kt’, the Fourier’s law of conductivity is expressed as follows: q = −kt∇T

(1)

In a particular area, the divergence of heat flux is given by the difference between the power generated in that region and the time rate of change in the heat in that region. Thus, it can also be represented as follows, ∇ · q = −kt∇ · ∇T = −kt∇ 2 T = g(r, t) − ρc p (∂ T (r, t)/∂t)

 (2)

where ‘r’ represents the coordinate in the region where the temperature is measured, ‘t’ corresponds to the time measured in seconds, ‘g’ is the power density per unit volume expressed in W/m3 , ‘cp ’ denotes the heat capacity measured in J/(kg K), and ‘ρ’ denotes the density expressed in kg/m3 . Thus, the Eq. (2) may be rearranged to form the following equation for heat transfer:

Thermal Analysis of Different Components …

ρc p ∗ (∂ T (r, t)/∂t) = kt∇ 2 T (r, t) + g(r, t)

149

(3)

The proportionality constant which is the thermal conductivity is isotropic in a uniform medium and for materials used in chip manufacturing, which are silicon and silicon dioxide for transistors, and the materials used for metal interconnects, which are aluminium and copper, have fundamental material properties in the standard tables where the values of thermal conductivity can be determined. The transient thermal response is described by the solution to Eq. (3). The derivatives of all the equations are taken with respect to time in the steady state with result to zero, so steady state analysis is equivalent to solution after solving the partial differential equation which is given by following equation called the Poisson’s equation. ∇ 2 T (r ) = −g(r )/kt

(4)

The thermal analysis starts by taking the input file after the complete placement. Then, the component properties are defined which are present on the PCB. Next, the PCB board to be used is selected, whether it is an IC, ASIC or FPGA, and accordingly, the board properties are defined. Now, the thermal analysis of the board is performed by using the ANSYS tool, depending upon the results the placement of the components on the PCB is changed and the thermal analysis is performed again until the desired output is achieved. The thermal analysis steps are shown in the form of a flow chart in the Fig. 1. Fig. 1 Flowchart of thermal analysis

150

K. Sanjitha et al.

3 ANSYS Tool for Thermal Heat Simulation John Swanson formed ANSYS in 1970, and during the 2000s, the company acquired a number of independent engineering design firms. The ANSYS software package is used to build models that measure a product’s stability, temperature distribution, fluid motions, and electromagnetic properties, as well as to design products and semiconductors. A division of the ANSYS tools which is the ‘electronics solutions’ is used by engineers to create innovative and advanced technology electronic products which are faster and very much advanced performance than the existing devices. ANSYS tools can be used to develop improved products, but for advanced digital projects, it is critical that the printed circuit board (PCB) packages are thoroughly evaluated using a trusted tool—simulation tool like Ansys SIwave-DC. Ansys SIwave is a specialised architecture tool for analysing electronic packages and PCBs for power integrity, signal integrity, and EMI. SIwave-DC, SIwave-PI, and SIwave are the three advanced research packages available.

4 Simulation and Results This paper uses the ANSYS tool to conduct a steady state thermal analysis of the PCB board in order to obtain the best arrangement of electronic components while keeping the total temperature of the PCB almost uniform, ensuring the device’s stable operation. In the first case, let us consider four square blocks of same size, but having different power dissipations to be placed on the PCB board. The temperature of Block A is 45 °C, Block B is 100 °C, Block C is 65 °C and Block D is 85 °C as shown below in the Fig. 1, and the PCB is kept at room temperature. After performing the steady state thermal analysis to the Fig. 2, the result is as shown in the Fig. 3. As shown in Fig. 3, the temperature distribution of the PCB is not uniform and it is observed that the temperature is more towards the right side and less towards the left, that is the temperature is increasing as we move from left to right, which causes a potential differences in the PCB which is undesirable and can cause some second order effects in transistors and would therefore affect the performance on the transistor which in turn effects the performance of the IC. Now, to resolve the drawbacks the electronic blocks A, B, C and D on the PCB are rearranged and steady state thermal analysis is performed a number of times until a uniform temperature all over the PCB is achieved. The final arrangement of the blocks on the PCB where the uniform temperature is achieved is as shown in Fig. 4. The blocks are rearranged such that the highest power dissipating block A and the lowest power dissipating block B are placed at close vicinity such that the overall temperature on the PCB is uniform. The simulated result for the arrangement in Fig. 4 is shown in Fig. 5. As shown in the Fig. 5, the temperature distribution on the PCB is uniform with the average temperature of about 80 °C.

Thermal Analysis of Different Components …

Fig. 2 Initial position of the square blocks of the same size

Fig. 3 The simulation result for the arrangement in Fig. 2

151

152

Fig. 4 Rearranging the blocks in the PCB to achieve uniform temperature

Fig. 5 The simulation result for the arrangement in Fig. 4

K. Sanjitha et al.

Thermal Analysis of Different Components …

153

Fig. 6 The blocks of four different sizes and different temperatures placed on the PCB

Now, consider a second case in which there are four rectangular blocks of different sizes and different power dissipations are taken on the PCB as shown in the Fig. 6. The thermal analysis is performed, and based on the analysis results, the blocks are rearranged and the same procedure is applied as discussed earlier. After a few experimental trials are performed, the final positions of the blocks A, B, C and D are achieved, which has uniform temperature distribution as shown in Fig. 6. The corresponding simulation results for the Fig. 6 are shown in Fig. 7. Similarly, in the third case five different blocks with different shapes and different power dissipations are considered as shown in the Fig. 8. There is a circular block with the highest power dissipation. After a few experimental trials are performed, the final positions of the blocks A, B, C, D and E are achieved which have the uniform temperature distribution as shown in Fig. 8. The corresponding simulation results for the arrangement in Fig. 8 are shown below in Fig. 9.

5 Conclusion The thermal analysis is very essential in the placement of electronic blocks on the PCB to form a desirable design. This paper describes the thermal analysis of PCB; using ANSYS tool, the steady state and the following conclusions are obtained, • According to the simulated results obtained from the thermal analysis, it is understood that the arrangement of electronic components on the PCB board can affect

154

Fig. 7 The simulation result for the arrangement in Fig. 6

Fig. 8 Five blocks of different shapes and different power dissipations

K. Sanjitha et al.

Thermal Analysis of Different Components …

155

Fig. 9 The simulation result for the arrangement in Fig. 8

the temperature distribution and the performance of the system is affected. Therefore, to minimise the temperature in the PCB, the components should be located in a suitable location based on their properties. • The electronic components with more power dissipation should be placed near the corners of the PCB depending upon the components connectivity and should also be placed away from temperature-sensitive and heat-sensitive components, as it may affect their results. Thus, it can be concluded that steady state thermal analysis is an important and using the ANYSYS tool for analysing the thermal stability of PCBs can help minimise construction costs. It can also help with PCB structure design.

References 1. Funk JN, Menguq MP (1992) A semi-analytical method to predict printed circuit board package temperatures. IEEE Trans Compon Hybrids Manuf Technol 15(5) 2. Ismail FS, Yusof R (2009) Thermal optimization formulation strategies for multi-constraints electronic devices placement on PCB. In: Proceedings of TENCON 2009–2009 IEEE Region 10 conference 3. Narayana VA, Premchand P, Govardhan A (2009) A novel and efficient approach for near duplicate page detection in web crawling. In: 2009 IEEE international advance computing conference, IACC 2009 4. Zhan Y, Kumar SV, Sapatnekar SS (2007) Thermally aware design. In: Foundations and trends R in electronic design automation, vol 2(3), pp 255–370

156

K. Sanjitha et al.

5. Li RS, Larson SE (2000) An approach to thermal analysis of PCB with components having cyclic electrical loading. In: The proceedings of ITHERM 2000 conference 6. Devadasu G, Sushama M (2016) A novel multiple fault identification with fast fourier transform analysis 1st international conference on emerging trends in engineering, technology and science, ICETETS 7. Joshi Y (2002) Emerging thermal challenges in electronics driven by performance, reliability and energy efficiency. In: 8th international workshop on thermal investigations of ICs and systems (THERMINIC) 8. Li P, Pileggi LT, Asheghi M, Chandra R (2004) Efficient full-chip thermal modeling and analysis. In: Proceedings of the IEEE/ACM international conference on computer-aided design, pp 319–326, Nov 2004 9. Ahmed SM, Kovela B, Gunjan VK (2020) IoT based automatic plant watering system through soil moisture sensing—a technique to support. Adv Cybern Cognit Mach Learning Commun Technol 643:259 10. Chang H, Sapatnekar SS (2007) Prediction of leakage power under process uncertainties. In: Proceedings of the ACM transactions on design automation of electronic systems, vol 12(2), Apr 2007 11. ANYSYS website: https://www.ansys.com/en-in/products/photonics/heat 12. Gunjan VK, Pathak R, Singh O (2019) Understanding image classification using tensor flow deep learning-convolution neural network. Int J Hyperconnectivity Internet Things (IJHIoT) 3(2):19–37 13. Cong J, Luo G, Wei J, Zhang Y (2007) Thermal-aware 3D IC placement via transformation. In: Proceedings of the Asia-South pacific design automation conference, pp 780–785 14. Dadgour H, Lin S-C, Banerjee K (2007) A statistical framework for estimation of fullchip leakage-power distribution under parameter variations. IEEE Trans Electron Devices 54(11):2930–2945 15. Ahmed SM, Kovela B, Gunjan VK (2021) Solar-powered smart agriculture and irrigation monitoring/control system over cloud—an efficient and eco-friendly method for effective crop production by farmers in Rural India. In: Proceedings of international conference on recent trends in machine learning, IoT, smart cities and applications, pp 279–290. Springer, Singapore

Custom Handloom Mobile Application K. Pushpa Rani, N. Sushma swaraj, K. Himabindu, S. Shashank, and B. Vara Prasad

1 Introduction Handloom products are traditional textile art for clothes. The production of handloom products is most important for economic development in rural India [1]. From which the handloom saris are the most producing product among all. Several regions have their traditions of handloom sari. From every state, we have different types of fabrics and designs for handloom products. Every design has a unique feature for the product. Nowadays, everything is available online. Our project custom handloom mobile application is an online shopping application for the customization of our products. Nowadays, every product is customized as we wish. Handloom products are the traditional attires from every place [2]. We can have many products like saris, bags, doormats, and many more. From which handloom saris are very different from every area. We have different designs, and each of them explains differently [3]. Our project is about finding the best product we need to have of handloom made. Buying handloom products can also encourage the production of rural people [4, 20]. It tells us about our culture and tradition. In our project, we are using the customization option to make consumers satisfied. So we want to try to source products from local artisans and designers as much as possible and create/promote their brand, giving a boost to micro- and small units [5]. We are using the Android studio tool to build our project. Email-id and phone number are a must for log-in to the application [6, 19].

K. Pushpa Rani (B) · N. S. swaraj · K. Himabindu · S. Shashank · B. V. Prasad (B) Department of Computer Science and Engineering, MLR Institute of Technology, Hyderabad, India e-mail: [email protected] B. V. Prasad e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_18

157

158

K. P. Rani et al.

2 Literature Survey There are some existing methods [7, 8] for handloom products like (1) OdiKala (2) National Handloom Corporation (3) Weavesmart A.

OdiKala:

OdiKala discovers unique products of Odisha. It includes handloom, handicraft, ethnic, organic, and natural products from Odisha. Which, let the whole world explore Odisha.OdiKala.com is the first step toward this vision [9]. This application provides only the handloom products of Odisha. In this application, we do not have any custom options. It is a handloom shopping application. It does not provide all the types of handloom products as it only has Odisha products. B.

National Handloom Corporation:

National Handloom Corporation is a one-stop supermarket chain that aims to offer customers a wide range of home and personal products under one roof. Each store has a stock of home utility products like beauty, food, garments, kitchenware, and more [10, 17]. It has many products but we need to shop it from the store. This application gives us only information about their shops. And for some products, we do not have price value. We need to buy the product by going to the shop which they have given the location details [11–13]. C.

Weavesmart:

Weavesmart’s goal is to bring together the weaving community and the buyers into close contact. That is to connect weavers and wearers [14, 18]. They are having different types of saris, dresses, fabrics, and many more. This is an online web application for different categories of handloom and silk saris, and more. But we do not have any custom options [15, 16].

3 Requirements A.

Hardware Requirements: • • • •

B.

Core I5 Processor Minimum of 4 GB RAM, 8 GB recommended Minimum of 40 GB Hard Drive Android Phone

Software Requirements: • Operating System: Windows 7/8/10 • Development Tools: Android SDK, Android Emulator, Android Studio • Database Platform: Firebase

Custom Handloom Mobile Application

159

4 Proposed System After farming, handloom weaving is the most important livelihood in rural India. Improving handloom products will directly affect the lives of millions of Indians economically. We need to encourage traditional wear and the handmade products of our places. These products show our culture in it. So to do that, we are bringing this application with the customization option. The application is about handloom products where people can buy only handmade products. In this application, we are providing the custom option. The consumers can choose this option and can contact the maker to get a customized product. Customers can buy if they like the product which is already available in the application. Otherwise, they can get customized products. We provide a chat box where the consumer contacts the maker to make their design get done with them. The customer needs to register first into the application. Then they can browse for the product and selects the product which they want to buy. Then they can choose the option between getting customized or buying the original product. The maker also needs to register, and then they need to upload their products. When a customer wants to use the custom option, then the customer contacts the maker. They decide the design or what changes they need. They finalize the product after discussion. The custom option is used when the customer is willing to buy the product and can pay the extra charges. While placing an order, the customer can choose the payment option, either online mode or cash on delivery. When the customer receives the product, the acknowledgment is sent to their account. This is how the application works. A.

System Architecture:

The system architecture is the components that are the base of the application. In this architecture, we are using different components, which are shown in the diagram (Fig. 1). 1.

Customer or Designer:

The person who wants to buy the product should register as a customer and the person who wants to sell the product should register as a designer. 2.

Mobile Phone:

Mobile phone is required to use the application. As it is an android mobile application, the user needs to have an android mobile phone. 3.

API Gateway:

An API gateway is an API management tool. It helps to communicate between the client and the collection of backend services. It accepts all the application programming interface calls and returns the appropriate results on the screen to the client.

160

K. P. Rani et al.

Fig. 1 System architecture

4.

Online Application Platform:

The online application platform is nothing but the application process. We have different processes like account registration, product catalog, cart server, order server, and delivery. The customer needs to follow these steps to buy a product. 5.

Databases:

Databases are used to store data of the customer’s account, product catalog, cart, and order details. In this application, we are using firebase to store the data. For log-in authentication, to store data, firebase is used.

5 Results The above graph shows us the usage of the application. It is showing the results of the customers who created their accounts in our application. The graph changes according to the number of registration counts in the application. This graph is the sample results of our application (Figs. 2 and 3). The above graph is about the stages of different products where people view the product, add the product into the cart, buys the product, and give a review to the product (Fig. 4).

Custom Handloom Mobile Application

161

Fig. 2 Application usage graph

Fig. 3 Product stages

Product Stages Products View

Fig. 4 Types of product sales

Cart

Order

Review

Types of product sales Saree Dress materials Home Decor Fabrics

The above graph shows the percentage of different types of product sales where the saris have the highest production.

6 Conclusion The handloom application in this research contributes us to shop the handloom products online, which are the customized products. Consumer shows interest to make their customized products than the existing one. They can be unique from the others. The weavers also can improve their productivity by adapting to new trends day by day. They can make the product trendy, and also, it includes the traditional design. So, this research can help both of them. A consumer can buy the custom product if they want. Through this application, we can create our designed products. But the customer should be aware of the fabric so that the design can be a workout. The customer and maker can communicate and then finalize the product. This helps the

162

K. P. Rani et al.

consumer as well as the designer to make the products as they want. This also can help to encourage the handloom products and can improve economic growth to the weavers. Serving the customers with the original and the best-customized products and making them happy is the aim of this research. So, we hope that this research will have an effective way for the customers.

References 1. Annapurna M, Bijker WE (2018) Innovation in Indian handloom weaving. Technol Culture 59(3):509–545. https://doi.org/10.1353/tech.2018.0058 2. Sree Vyshnavi VP, Nair SS (2017) Handloom sector in India: a literature review of government reports. 4(8) 3. Pushpa Rani K, Reddy M, Anjaneyulu B, Hari Chandana B (2019) A formal assessment paper on EDUSET-MBA universities custom search. Int J Innov Technol Exploring Eng 9(1). ISSN: 2278-3075 4. Merugu SS, Kumar V (2020) Augmented reality on sudoku puzzle using computer vision and deep learning. Advances in cybernetics, cognition, and machine, 2020, Lecture Notes in Electrical Engineering, Springer, Singapore, pp 567–578 5. Paul R, Goowalla H (2018) A study on consumer awareness of handloom products with special reference to Dimapur District-Nagaland. 2(3) 6. Lakshmi L, Pushpa Rani K, Purushotham Reddy M (2019) A comparative study of navigation techniques and information retrieval algorithms for web mining. Int J Adv Trends Comput Sci Eng 8(1.3):10–14 7. Dissanayake DGK, Perera S, Wanniarachchi T (2017) Sustainable and ethical manufacturing: a case study from handloom industry. https://doi.org/10.1186/s40689-016-0024-3 8. Goswami R, Jain R (2014) Strategy for sustainable development of handloom industry. 6(2):93– 98. ISSN 0975-6477 9. Odikala: Online Shopping Ikat Handloom Clothes of Odisha. Site: odikala.com 10. National Handloom Corporation, Available: nationalhandloomcorp.com 11. Amaravathi G, Bhavana Raj K (2019) Indian handloom sector–a glimpse. 8(6S4). ISSN: 22783075 12. Pushpa Rani K, Jhansi M, Chandrasekhara Reddy T (2011) Best keyword cover search using keyword-nne algorithm. Int J Mech Eng Technol 8 13. Pushpa Rani K, Roja G, Sabitha C, Dhana Lakshmi B, Sreeja S (2018) A new approach for converse binary tree traversals. Int J Eng Technol (UAE) 7(4) 14. Weavesmart-Largest handloom display for silk sarees. Available: weavesmart.com 15. Mishra V, Shivakumar S, Ravishankar VD, Bhalla K, Dhiman B (2019) Design and development of Khadi-Kart: a web-based application for rejuvenating the Handloom Industry in India. https:// doi.org/10.1109/GHTC46095.2019.9033041

A Compact Orthomode Transducer in K-Band for Satellite Communications V. Santhosh Kumar, NageswaraRao Lavuri, and Abdul Subhani Shaik

1 Introduction The demand for innovation and testing services is improving everyday, and several implementations in satellite systems, navigation technology, and data perceiving involve the development of modern and reliable systems and devices. Bandwidth enhancement in a desired band is a significant variable or industry as well as productivity in the microwave territory. One of the most utilized band for this technology is K-band. Waveguide in a component is a key factor in interpreting a recurrence utilized. Waveguide innovation is the conceptual model of radiator feed systems. Incorporate a wide range of elements to establish a microwave sub-components [1] and evolve their variables to adequately direct the wave. There have been internal and external microwave modules in its most prevalent design. The intended system review describes a device utilized in a reflector feeding system as an orthomode transducer depicted in Fig. 1. Ortho-mode transducers (OMTs) were widely utilized when double polarizations are involved in radio wire feeding structures [2]. Generally, they can be utilized to acquire double-polarized signals via horn component also splitting them through two discrete terminals. Broadband fidelity is commonly done using symmetrical designs [3], such as the Bifot crossroad and turnstile convergence [4]. In recent, with the development of communication implementations, double orthogonal polarization [5] transmissions have been desired to be obtained concurrently via a single antenna encompassing the intended band. An implemented specified band OMT is appropriate. V. S. Kumar (B) · A. S. Shaik Department of ECE, CMR College of Engineering & Technology, Hyderabad, India e-mail: [email protected] N. Lavuri Department of ECE, CVR College of Engineering, Hyderabad, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_19

163

164

V. S. Kumar et al.

Fig. 1 OMT schematic diagram

In view of the most recent structure and machining methods [6], functioning components were evolved at various frequencies and afforded a fascinating adaptability in terms of microwave variables and technical restraints. A circuit has been implemented to describe the projected behavior of a feed and optimal elements with a reliable device construction, simulation, and interpretation. CST is focused on the 3-D electromagnetic simulation. By using CST tool, the various microwave components are analyzed and parameters like return loss, isolation, and insertion loss are characterized. These methods dynamically adjust and reveal a quite good grasp of calculated theory and simulation results. An implemented OMT demonstrates how a K-band OMT with WR-42 port is constructed, simulated, customized, and evolved as receive ports. In the flared waveguide segment, the coupling aperture was already customized for attaining port-to-port isolation more prominent than 32 dB and creating it compact.

2 OMT Design The OMT is a waveguide junction with two aligned ports of 90° that isolate orthogonal polarizations. Figure 1 exhibits the implementation of outlined structure. An OMT consists of 4 ports, however, only three main ports. The merge of the typical port would have structural alignment for signals working in the same group, commonly circular or square. From the above outline of the OMT, the optimal OMT scatter matrix would be provided in Eq. 1. ⎡

0 ⎢0 S=⎢ ⎣0 e j∅1

0 0 e j∅2 0

0 e j∅2 0 0

⎤ e j∅1 0 ⎥ ⎥ 0 ⎦ 0

(1)

A Compact Orthomode Transducer in K-Band …

165

In certain expressions, the attachment between the direct (coupling) port and the vertical (horizontal) polarization at a common square port should have magnitude of 1 and a phase, and the return loss of all ports must be optimal; i.e., the reflection coefficient of each port should be zero. The K-band OMT design and interpretation were described. The OMT is utilized in K-band chain to isolate the symmetrically polarized signals within the same recurrence band. This design depicts the K-band orthomode transducer (OMT) construction and advancement. It comprises of the common throughput of portion waveguide [7], and a common port to rectangular waveguide, straight waveguide section (WR42), and an orthogonally connected port (WR42). A septum polarizer response is aligned to an OMT input. To attach an analogous RF input to a prevalent channel, a flared waveguide segment is employed. An intended OMT does have a substantial port and associated channel with the intended range. Revealed structure deploys common waveguide like a prevalent channel to concurrently acquire vertical as well as horizontal polarizations. A spherical stepped waveguide connector that offers a transition from the common channel to the fixed WR-42 functionality accomplishes tracking to axial port. Enhancing the network capacity and increasing a transversal channel matching with utilizing further stages is achievable. A lateral port section is vertical to a longitudinal hub or is in the point of convergence for a symmetrical structure. This resonance ailment affects in the termination of perceptual modes emitted in the common port for H-polarization input [8, 9]. This keeps the parallel port from linking with the better execution forms. The composition of OMT is depicted in Fig. 2, comprised of an input port (3), discrete intersection polarization, as well as 2 output sections (1 and 2) of each polarization. Ports 1 and 2 are identical WR-42 (Ld × Wd – 10.6 mm × 4.318 mm) rectangular waveguides, and port 3 is a square waveguide (L × W – 10.6 mm × 10.6 mm). The coupled port is located at an offset range of 12.256 mm, orthogonal to the transverse line also spherical to this. Figure 2 demonstrates the perspective view and cross-sectional view of the optimized OMT. The length and width of an associated space are seen as 5.942 mm and 2.514 mm consequently, and also the 1 mm span screw with a depth of 0.75 mm is embedded on a square port waveguide with a 3 mm counterbalance. The optimized dimensions of the OMT are L1 = 13.6 mm, L2 = 15.428 mm, L3 = 5.1 mm, and L6 = 12.256 mm. The coupled port has the dimension of LS = 5.582 mm, WS = 2.514 mm, L4 = 6.628 mm, and L5 = 6.7 mm. An intended structure attains the specified K-band frequency range. An OMT simulation is performed in the K-band. The CST tool [20, 21] has been utilized to optimize and simulate the OMT.

3 Simulated Results and Discussions The intended OMT is analyzed and simulated. Reflection coefficient of the intended model is exhibited in Fig. 3. The configured orthomode transducer accomplishes the K-band recurrence ranges The reflection coefficients (S11 & S33) of K-band

166

V. S. Kumar et al.

Fig. 2 Designed model a Perspective view of OMT; b OMT 3-D view

(24.5–25.4 GHz) in direct port and common port are better than—20 dB as well as reflection coefficient (S22) of at coupled port is –12 dB. An intended structure is reasonable for K-band. Figure 4 demonstrates the simulated loss of insertion in both polarizations. The

A Compact Orthomode Transducer in K-Band … Fig. 3 Simulated return loss in K-band

Fig. 4 Simulated insertion loss a S31, b S32

167

168

V. S. Kumar et al.

computed insertion loss (S31) is approximately 0.06 dB throughout desired frequency range also insertion loss (S32) is about 0.35 dB across an intended frequency. Figure 5 illustrates an isolation of configured system. An isolation is described as propagation in between the ports rectangular waveguide. An implemented structure has isolation greater than 32 dB across desired K-band. An isolation of the intended configuration is reasonable. Figure 6 demonstrates the VSWR of the recommended model. VSWR of K-band (24.5–25.4 GHz) at port 1 is about 1.22 over a desired frequency range. Figure 7 illustrates the VSWR of port 2 of intended structure. Simulated VSWR in port 2 is less than 1.8 over a specified band. VSWR of K-band at common port 3 is presented in Fig. 8. The intended model has VSWR about 1.2 across frequency range.

Fig. 5 Simulated isolation in K-band

Fig. 6 Simulated VSWR at port 1 in K-band

A Compact Orthomode Transducer in K-Band …

169

Fig. 7 Simulated VSWR of port 2

Fig. 8 Simulated VSWR at port 3

4 Conclusion In this paper, a compact OMT for waveguide feed networks has been designed and simulated. To fulfill the prerequisites of quintessential K-band applications, the compact OMTs have been configured. The configured OMT includes the frequency range of 24.5–25.4 GHz for coupled port and a direct port. Discovered structure has isolation greater than 32 dB, and also, all ports’ reflection coefficient is better −12 dB across the working frequency range. An enacted configuration has insertion loss on a desired frequency band around 0.35 dB. This design is implemented for the task of a K-band that can be utilized in satellite customized structures. The architecture is easy without sophisticated grooves and empowers robust appearance to be accomplished.

170

V. S. Kumar et al.

A modular OMT is easy to acquire, and OMT will have a size limit. The designed OMT is suitable for satellite systems.

References 1. Amyotte E,Demers Y, Hildebrand L, Forest M, Riendeau S, Sierra-Garcia S, Uher J (2010) Recent developments in Ka-band satellite antennas for broadband communications. In: AIAA international communications satellite systems conference 2. Boifot AM, Lier E, Schaug-Pettersen T (1990) Simple and broadband orthomode transducer. Proc IEE 137(6):396–400 3. Ruqiya CGK, Nageswara Rao L (2019) Design of Te21 mode coupler in Ka band for antenna feeds. Int J Innov Technol Exploring Eng 8(10):4539–4542 4. Uher J, Bornemann J, Rosenberg U (1993) Waveguide components for antenna feed systems: theory and CAD. Artech House, Norwood, MA 5. Pisano G, Pietranera L, Isaak K, Piccirilo L, Johnson B, Maffei B, Melhuish S (2007) A broadband WR10 turnstile junction orthomode transducer. IEEE Microw Wirel Compon Lett 17(4):286–288 6. Tribak A, Cano J, Mediavilla A, Boussouis M (2010) Octave bandwidth compact turnstile-based orthomode transducer. IEEE Microwave Wirel Compon Lett 20(10):539–541 7. Tao Y, Shen Z, Liu G (2011) Dual-band ortho-mode transducer with irregularly shaped diaphragm. Progress Electromagnet Res Lett 27:1–8 8. Tao Y, Shen Z (2009) Broadband substrate integrated waveguide orthomode transducers. J Electromagnet Waves Appl 23(16):2099–2108 9. Narayanan G, Erickson NR, Grosslein RM (1999) Low cost direct machining of terahertz waveguide structures. In: Tenth international symposium on space terahertz technology, pp 518–528, Mar 1999

A Review on Stock Market Analysis Using Association Rule Mining R. Venkateswara Reddy, K. Venkateswara Rao, M. Kameswara Rao, and B. P. Deepak Kumar

1 Introduction In the stock exchanging market, income improvement rate is a crucial pointer for investors to foresee the creating firms later on. The creating firms remain for particular firms that can advance make and more get expands benefit per offer [1]. Stock market recommender system considering association rule mining (ARM) that endorses a course of action of stocks. The objective of this recommender structure is to help stock market investors, self-contained financial authorities and resource boss in their decisions by proposing enthusiasm for a get-together of significant worth stocks when strong affirmation of possible advantage from these trades is open [2]. Stock rundown checking is one of the noteworthy activities of cash-related firms and private examiners in settling on wander decisions. Since stock markets are a brain-boggling, transformative, and nonlinear component system its estimate is considered as a testing errand [3], stock is the most popular of cash-related business segment instruments. It can be portrayed as a sign of capital enthusiasm of a man or an endeavor in an association or a limited hazard association [4]. Stock expenses depend upon various parts, the basic ones being the market notion, execution of the business, acquiring comes about and anticipated compensation, takeover or merger, presentation of something else or presentation of a current thing into new markets, share repurchase, disclosures of R. V. Reddy (B) · K. Venkateswara Rao Department of CSE, CMR College of Engineering and Technology, Kandlakoya(v), Medchal, Hyderabad, Telangana 501401, India e-mail: [email protected] M. Kameswara Rao Department of Electronics and Computer Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, Guntur District, A.P, India B. P. Deepak Kumar Department of CSE, CMR Technical Campus, Kandlakoya(v), Medchal, Hyderabad, Telangana, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_20

171

172

R. V. Reddy et al.

advantages or rewards, expansion or discharge from the summary [5]. Stock market expectation has been a district of uncommon eagerness in light of the capacity of getting an excellent yield on the place trade out a short time span [6]. The straightforward accessibility to databases with recorded costs influences the weak structure to market viability test a less requesting endeavor to realize differentiated and the profitability trial of the other two structures [7]. The divulgence of particular illustrations, which are trademark for time game plan and thusly have an extraordinary noteworthiness, is a basic errand in the area of time course of action examination, and similarly, the disclosure of certain case, to be particular of association rules, is a fundamental task in the examination of significant worth based data [8]. To envision peril and return factors exactly, the effective segments ought to be recognized. Honestly, one of the key issues of stock desire plot lies on the most ideal approach to pick operator features for gauge [9]. Transformative algorithm generally creates individuals with a particular true objective to fabricate their wellbeing, and it secures the perfect or close perfect individuals which address action rules when they are used as a piece of expert control structures [10]. Stock market is a dynamic and complex structure with uproarious, non-stationary data, making the conjecture of future stock costs one of all the more troublesome issues [11]. In the extent of creating markets, China is indeed unique from various perspectives and worth the effort of correct work both for its own motivation (it has various extraordinary components) and for the light it can hurl on the relationship among adequacy and business pa2 headway. The Chinese stock market has existed for barely 10 years and is winding up outstandingly rapidly [12]. Association rule mining is a defender among the most invaluable data mining frameworks whose standard point is to focus associations in value-based databases among sets of items or objects [13]. Techniques of association govern mining can be used to discover darken or covered association between things found in the database of transactions [14]. Association administer finds connection between evidently disengaged data among a far-reaching course of action of data things. Along these lines, connection guidelines can basically find a potential relationship among segments and pass on shock to us. One of the algorithms used for visiting thing sets mining and association run learning is Apriori algorithm [15]. The paper was one of the pioneers in use of the association rules to determine if there existed any relation between stocks accumulated under different sectors [16]. Existing investigates give a lacking point of view of the thought coast issue in stock market as a result of controls in examination approach and component decision [17]. The inspiration driving the examination is to check and analyze association stock esteem advancements by executing the association manage mining algorithm to focus norms of connection between improvements of association stock expenses from time to time [18]. To update the prescient influence of the fiscal time plan models different computational information-based techniques like the artificial neural network, fluffy induction framework, support vector machine, relevance vector machine hybridized with the cash-related time course of action models have been proposed in the literature [19]. Soft registering strategies are for the most part associated with stock market issues. They offer important contraptions in suspecting uproarious circumstances like stock markets, getting their non-straight behavior. In later past counterfeit honeybee settlement algorithm have been created

A Review on Stock Market Analysis Using Association Rule Mining

173

to demonstrate the insightful practices of bumble bee swarms and associated for settling combinatorial sort problems. Artificial bee colony (ABC) is an oddity in the district of algorithms advancement, energizing premium and preferable outcomes pondered over various algorithms, as genetic algorithms (GA) and particle swarm optimization (PSO), in a couple of regions of examination. Therefore, pick the ABC to find the best course of action of slacks. 19 monetary examiners who give in stock markets for the most part are not cautious of the stock market execution. The issue of stock exchanging is being confronted, because they do not know which stocks to be purchased and which to sell keeping in mind the ultimate goal of receiving benefits. These sorts of issues are settled by making ideal stock rules in the stock market forecast. In this paper, for producing ideal stock rules to help the stock market forecast by utilizing associate rule mining algorithm.

2 Literature Survey Among the most essential field of data mining, mining association rules is exceptionally good. Mining association rule can be used when we want to know the pattern exhibited by customer or any other things. Mining association rules, are called as market holder analysis, which is a key while doing data mining. While concentrating on the purpose of offering exchange data, it is important to look at the consumer buy behavior and help with expanding the deals and ration stock. This fills the scientists in as a large region to develop a superior data mining algorithm. This section addresses a description of the current rule mining of associations. Review on Association Rule Mining Algorithm Market basket analysis is one of the use cases for association rule mining algorithm, and this is majorly used to recommend products to retail customers by analyzing the past behavior of same customer or any other customer with similar characteristics. These rules majorly focus on identifying what products are usually bought together and then accordingly recommend products to customers. Imagine a store environment, all the customer transactions are saved in the database which will have what products are bought together by this customer. The arranging division may be concerned with finding the “associations” among different items with pre-determined certainty based on past behavior. These kinds of associations may be useful in the preparation of advancements and rebates or retire association and store format. The yield of the examination frames the allowance for suggestion or marketing systems. In all associations lead mining algorithms, the most widely recognized advancement is to section the activity into two sub-assignments. Rule age: The sub-undertaking separates all the high certainty rules from the continuous item sets acquired in the above advance. These rules are called solid rules. Zhixin et al. proposed an improvised order system in view of predictive association rules. Grouping-dependent predictive association rules (CPAR) is one of the ways of

174

R. V. Reddy et al.

association characterization technique which includes the advantages of acquainted order and the regular govern-based arrangement. For the age of the manager, CPAR is more efficient than the ordinary administer-based order, because many of the repeat computations are skipped and various literals can be selected to make numerous rules in the meantime. Despite the fact that the advantage said above stays away from the duplicate estimation in run age, the expectation forms have the detriment in class manage dispersion irregularity and interference of incorrect class rules. Also, it is not much efficient in the cases where it fulfills no rules. The creator suggests class weighting adjustment, center vector-based cantered pre-grouping, and post-preparing along with support vector machine (SVM) to stay away from these difficulties. In the classification association policy mining, Wang et al. recommended a new weighting approach to administration. Characterization association rule mining (CARM) is the most up-to-date order to determine mining strategy by using classification association rules (CARS) to create an association to administer miningbased classifier. The specific CARM algorithm is often not respected even if it is used, a comparable CARs structure can be consistently generated using the data, and a classifier is usually given as an organized CAR list, contingent upon chose lead requesting method. In the current past, a few control requesting approaches have been received, which can be organized as run weighting, bolster certainty and half and half. In this method, an elective administers weighting strategy, known as class item score-based rule weighting (CISRW) and a manage weighting-based lead component relying upon CISRW. After that, two cross-breeds can be integrated and developed by consolidating support-certainty and CISRW. Zongyao et al. suggested a mining neighborhood association design from spatial dataset. The designer suggests a model and a method to mine nearby associations from the present spatial data, while taking account of the fact that spatial heterogeneity can be commonly expressed in validity. The model’s critical component is: algorithm localized measure of association strength (LMAS), which is used to measure the design of neighborhood associations. Spatial association relations can be completely represented as spatial relationships which are demonstrated by DE-SIM model. Also, the developer offers a way for mining to choose nearby association designs that are taken from the same type of spatial data. Mines reference procedure and target objects mostly have association examples and procedures LMAS for every question in the reference chosen object for some intrigued spatial connection. In this way, the impact of the algorithm is a LMAS conveyance delineate rehashes association potential varieties within the tested region. Spatial insertion for LMAS is recommended to deliver a consistent LMAS circulation, which is used to look at problem areas that show solid association designs. This procedure was connected in an environmental framework inquire about. Yong et al. recommended a different mining association govern by another measure criteria. Nowadays, association rules mining from huge databases is an interesting area of research taken after by numerous application regions. Then again, there are a few troubles in the solid association rules mining, contingent upon the help certainty system. Right off the bat, there are countless association rules produced, and it is then entangled for the client to locate the intriguing ones. The link alongside

A Review on Stock Market Analysis Using Association Rule Mining

175

the highlights of the predetermined application zones is then skipped. Imaginative measure parameter like chi-square test must begin association rule mining. We can also minimize the measure of rules using this chi-square test. The front of measure and the chi-square test are used by the creator for association rules mining to use item sets that are not measured, as regular item sets or rules are made. Hence, the quantity of examples of item sets is limited and is also easy to the client to comply with very good observable association rules. Chi-square test is compelling to reduce the number of examples by mixing backing and cover oblige. As per the chi-square test, minimal insignificant qualities are omitted, and hence, the productivity and mining association rules reality are enhanced. Vo et al. gave mining customary association rules using successive item sets cross section. A numerous technique has been computed for the changing the time in mining incessant item sets. However, the techniques which have to deal seasonality of mining association rules are often not considered for in depth research. Really, season of mining association rules is substantially more prominent under the database that consists of various incessant item sets (from 10 thousand up to millions) than the one needed for mining successive item sets. An application for grid in mining ordinary association rules is created which will significantly reduce the ideal opportunity or the rules of mining. This procedure has two phases (1) development of regular item sets cross section and (2) mining association rules from grid. The parent-kid connection in grid is used for the quick assurance of association rules. Rastogi et al. proposed mining upgraded association rules with absolute and numeric traits. Mining association rules on immense data have recently accomplished a greater consideration. Association rules are robust for predicting relationships including highlights of a connection and incorporate various use cases in marketing and retail areas. Advanced association rules are great way to deal with given importance to the most important highlights interfacing few given properties. Improved association rules are rational to include incomparable characteristics and also unpredictable to find those where management aid or certainty is extended. Here, the difficulty of the improved association rules is disentangled in three ways to be specific, 1. 2. 3.

Association rules allow for the inclusions of disjunctions over incomparable highlights. Association rules allow for the inclusion of an irregular number of incomparable highlights. Incomparable highlights can be either absolute or numeric. The general association rules lead to mine occasional and nearby examples interfacing different highlights. It likewise proposes a decent technique for pruning the hunt space while defining enhanced association rules for both straight out and numeric highlights.

Wang et al. played out an examination on association rules mining in light of chronology in online business. Business activities done with the utilization of the Internet are exceptionally well known. A lot of exchange logs are created, which gathers valuable data by using data mining. Thus, association rule mining is critical

176

R. V. Reddy et al.

for online business. In any case, there are a few issues that emerge in the current association rules mining frameworks. The current regular systems cannot take care of these issues. The objective is—it proposes association rules mining based upon metaphysics. There are three sections of amid data mining: 1. 2. 3.

Techniques for metaphysics development and standards of product classification; R-interesting simplification based on actual situations; Association rules mining execution based upon metaphysics using enhanced Apriori. It also tests the enhanced algorithm by making use of Food Mart2000, Java used as advancement dialect and Jena as philosophy, finishes the complete procedure of mining, and the legitimacy of algorithm by the checking.

3 Association Rule on Stock Market Agrawal et al. [3] finding associating rules is an important issue in data mining. There is significant research going on in utilizing association rules in the field of data mining. The association rule algorithm is used majorly to identify the connections between different transactions or highlights those transactions that happen synchronously in the database. Let us take an example: If a given customer purchases thing X, and also thing Y, then we can determine that there is a connection between thing X and thing Y. Identifying this connection or association is very important. Therefore, the main reason behind actualizing the association rules algorithm is to discover parallel connections by investigating random data and to use these findings regarding connections as a source of perspective amid basic leadership [3]. A standout among the most imperative issues in current back is finding productive approaches to outline and envision the stock market data to give people or foundations helpful Hajizadeh et al. 115 data about the market conduct for investment choices. The gigantic measure of profitable data created by the stock market has pulled in analysts to investigate this issue space utilizing diverse techniques. Shu Hsien et al. examined issues on investing in Taiwan stock market by utilizing a two-phase approach in data mining. Principal arrange Apriori algorithm is a philosophy of rules of association, which is actualized to mine information and represent learning examples and rules keeping in mind the end goal to propose stock class association and conceivable stock classification investment accumulations. At that point, the K-implies algorithm is a procedure of group investigation executed to investigate the stock bunch keeping in mind the end goal to dig stock classification bunches for investment data. Thus, they propose a few conceivable Taiwan stock market portfolio choices under various conditions.

A Review on Stock Market Analysis Using Association Rule Mining

177

4 Mining Various Kinds of Knowledge from Databases Data mining is an application subordinate issue, and various applications may require a distinctive mining system. When all is said in done the sorts of information which can be found in databases are classified as takes after: Mining Association Rules A gigantic measure of data is put away electronically in many ventures. Specifically, in all retail outlets the measure of data put away has become gigantically because of bar-coding of all products sold. For instance, Walmart, with in excess of 4000 stores, gathers around 20 million purpose of offer exchange data every day. Given this pile of data, it is great marketing prudence to attempt to investigate it to discover data. This is an interesting case of investigating a large database of supermarket exchanges with the point of discovering association run the show. This is called association rules mining or market basket analysis. It includes hunting down interesting client propensities by taking a gander at associations. Association rule mining has numerous applications other than market bushel examination, including marketing, client division, solution, electronic trade, grouping, bunching, web mining, bioinformatics, and accounts. Basic case of Transactions The retailer needs to know which things are sold together much of the time. We expect that the quantity of things in the shop stock is n, and these things are spoken to buy I {il, i2, … 1}. We indicate exchange by T {t1, t2… …} each with a one of a kind identifier (TID) and each determining a subset of things from the thing set I bought by one client. Every exchange of m things be {il, i2, ……, im) with m < n. Presently, find association relationship, given a large no. of exchanges, with the end goal that things that have a tendency to happen together are distinguished. It ought to be noticed that association rules mining does not consider the amounts of things purchased. Association rules are frequently composed as X + Y. X is regularly alluded to as the manager’s forerunner and Y as administers resulting. It demonstrates that exclusive X and Y have been discovered together oftentimes in the given data and does not demonstrate a causal relationship inferring that purchasing of X by a client causes him/her to purchase Y. Assume thing X and Y seem together in just 10% of the exchange yet at whatever point x shows up there is 80% possibility that y additionally shows up. The 10% nearness of X and Y together is called support or pervasiveness of the manage 80% possibility is called certainty or consistency of the run the show. Basically, the help and certainty are measures of the interestingness of the run the show. An abnormal state of help showed that the manager is visiting enough for the business to be interested in it. An abnormal state of certainty demonstrates that the lead is genuine frequently enough to legitimize a choice in light of it.

178

R. V. Reddy et al.

The certainty of these rules is gotten by separating the help for the two things in the lead by the help for the thing on the left-hand side of the run the show. The certainty of the four rules consequently is 54 = 75%, % = 75%, 3/3 = 100%, and 3/4 = 75% individually. Since every one of them has a base 75% certainty, they all qualify. Grouping Analysis The method of gathering unique things or physical things into classes of comparative items is known as unsupervised arrangement or bunching. Grouping helps build significant parceling of large organization of articles in view of a “partition and overcome” approach and disintegrates a large-scale framework as minor groups to rearrange characteristics and usage. Pattern-Based Similarity Search. This sort of database includes: financial database for stock value record, therapeutic database, band interactive media databases. While hunting down comparable example in a transient or spatial-fleeting database, two kinds of questions are normally experienced in different data mining activities: 1. Object-relative similitude question (that is… run inquiry or comparative question) in which search is performed on a gathering of articles to identify the ones that are within a client characterized separate from the identified protest. 2. All-pair closeness question (that is… spatial join) here the objective is to identify all the combination of components that are within a client determined separation from each other.

5 The Methodology of the Study Data mining technique is intended to guarantee that it prompts a steady model, which effectively tends to the issue it is intended to unravel. Different data mining philosophies were proposed to fill in as diagrams for how to compose the way toward social affair data, dissecting data, spreading comes about, actualizing results, and checking changes [9]. To assemble the model that examines the stock patterns utilizing the choice system, the Cross-Industry Standard Process for data mining (CRISP-DM) [13] is utilized. This approach was given in the mid-1990s by a European consortium of organizations to fill in as a non-proprietary standard process display for data mining. The given model comprises of accompanying six stages: • • • • •

Identify the goal of mining the stock prices. Study the data organization and gathered data. Prepare the data which is used as a part of the order demonstrates. Finalize the strategy to fabricate the model. Validate the model by using one of the outstanding assessment techniques.

A Review on Stock Market Analysis Using Association Rule Mining

179

• Use the model in the stock market to foresee the best move to be made, if it is offering or purchasing the stocks. • Determine the use case and target of building a model. Profit Association Rules Mining: S. Independent Stock # of No company price exchanges direction

Dependent company

Stock # of Confidence price exchanges direction

1

SUNTECK

491

Associated TCFC with

Decrease 276

56%

2

RELIGARE Decrease 527

Associated TCFC with

Decrease 293

56%

3

BAJAJ

Decrease 520

Associated RELIGARE Decrease 288 with

55%

4

UNITECH

Decrease 524

Associated TCFC with

Decrease 290

55%

5

TCFC

Increase

530

Associated RELIGARE Decrease 293 with

55%

6

UNITECH

Decrease 524

Associated RELIGARE Decrease 289 with

55%

7

SUNTECK

Increase

Associated RELIGARE Decrease 270 with

55%

8

SUNTECK

Decrease 488

Associated UNITECH with

Decrease 268

55%

9

HDIL

Decrease 505

Associated UNITECH with

Decrease 277

55%

10

RELIGARE Decrease 527

Associated UNITECH with

Decrease 289

55%

Increase

491

The primary association decision demonstrates that company SUNTECK associated with TCFC with 56% confidence, that means, if share cost of SUNTECK increases (1) at that point TCFC will likewise go low (I). What’s more, the eighth association decided that company SUNTECK associated with UNITECH with 55% confidence that if share cost of SUNTECK drops, at that point UNITECH will also drop. The above sort of 488 exchanges is comprised in the exchange table. In the wake of applying combined intra-inter-exchange approach with window length six, now we apply the Apriori on this prepared data and discover the association rules using the characteristics. Information Data: ID

DATE

BAJAJ

TCFC

RELIGARE

HDIL

UNITECH

SUNTECH

1

1/1/2008

1

1

1

1

0

0

2

2/1/2008

0

1

1

1

0

1 (continued)

180

R. V. Reddy et al.

(continued) ID

DATE

BAJAJ

TCFC

RELIGARE

HDIL

UNITECH

SUNTECH

3

3/1/2008

0

1

0

0

1

0

4

4/1/2008

0

0

0

1

0

1

5

7/1/2008

1

1

1

0

1

0

…

….

….

…

……

…

……

…

981

28/1/2012

1

1

1

0

1

1

982

29/1/2012

0

1

0

1

1

0

983

30/1/2012

1

1

1

0

0

1

Profit Association Rules Mining: S. Independent Stock # of No company price exchanges direction

Dependent company

Stock # of Confidence price exchanges direction

1

HDIL

Decrease 501

Associated TCFC with

Decrease 292

58%

2

HDIL

Increase

Associated RELIGARE Decrease 268 with

57%

3

RELIGARE Decrease 524

Associated TCFC with

Decrease 292

56%

4

BAJAJ

Increase

456

Associated UNITECH with

Decrease 254

56%

5

UNITECH

Increase

454

Associated BAJAJ with

Decrease 252

56%

6

UNITECH

Increase

454

Associated TCFC with

Decrease 252

56%

7

BAJAJ

Decrease 517

Associated RELIGARE Decrease 286 with

55%

8

TCFC

Decrease 528

Associated RELIGARE Decrease 292 with

55%

9

TCFC

Decrease 528

Associated HDIL with

Decrease 292

55%

10

UNITECH

Increase

Associated RELIGARE Decrease 250 with

55%

472

454

The first association decision shows that HDIL and TCFC have 0.58 confidences; that is, if the share cost of HDIL decreases (↓) at that point TCFC will also decrease (↓). Also, the fifth association decision shows that UNITECH and TCFC have 68 certainty, that if share cost of UNITECH increases (↑) at that point TCFC will also decrease (↓). Consolidated inter-intra-exchange Algorithm with 8 as a sliding window length in the wake of applying a similar approach on exchange data with 8 as a sliding window length we discovered diverse outcomes.

A Review on Stock Market Analysis Using Association Rule Mining

181

Information Data: ID

DATE

BAJAJ

TCFC

RELIGARE

HDIL

UNITECH

SUNTECH

1

1/1/2008

1

1

1

0

1

0

2

2/1/2008

0

1

1

0

0

1

3

3/1/2008

0

1

0

1

1

0

4

4/1/2008

0

0

0

0

0

0

5

7/1/2008

1

1

1

1

0

0

…

….

….

…

……

…

……

…

981

28/1/2012

1

1

1

0

0

1

982

29/1/2012

1

0

1

0

0

1

983

30/1/2012

0

1

1

0

0

0

Produce Association Rules Mining: S. No

Independent company

Stock price direction

# of exchanges

Dependent company

Stock price direction

# of exchanges

Confidence (%)

1

SUNTECK

Decrease

485

Associated with

RELIGARE

Decrease

276

57

2

HDIL

Decrease

504

Associated with

TCFC

Decrease

268

57

3

UNITECH

Decrease

521

Associated with

RELIGARE

Decrease

293

56

4

RELIGARE

Decrease

525

Associated with

TCFC

Decrease

293

56

5

RELIGARE

Decrease

525

Associated with

UNITECH

Decrease

293

56

6

SUNTECK

Decrease

485

Associated with

BAJAJ

Decrease

269

55

7

BAJAJ

Decrease

518

Associated with

RELIGARE

Decrease

287

55

8

TCFC

Decrease

529

Associated with

RELIGARE

Decrease

293

55

9

HDIL

Decrease

504

Associated with

RELIGARE

Decrease

278

55

10

SUNTECK

Decrease

485

Associated with

TCFC

Decrease

267

55

Here we find that the length of sliding window means holes between the exchanges to discover inter-exchange rules. After a watchful perception, we could see that it is better to consider the most recent exchange for mining the rules on the grounds that the assurance is around meet while thinking about the window length. Hence, the sliding window length four, which is the best approach, is best of all.

182

R. V. Reddy et al.

6 Conclusion These variables are vital in market basket analysis because in connection with alternative techniques, one can arrive at a decision about, the performance of the proposed strategy, these are the important reasons for market changes in container examination development. These elements are speedily explained in the exploratory outcomes. The review analysis seems promising aftermath of that integrated intrainter-exchange approach. It thinks about the most recent exchanges to determine the rules, so it is better to pick least length window. It is better when there is immense measure of data, thereby decreasing preparing time leading to good certainty. It will be appropriate for forecasts and helpful in stock exchanging stages for legitimate investments in numerous segments and connection among numerous segments.

References 1. Alor-Hernandez G, Gomez-Berbis JM, Jimenez-Domingo E, Rodríguez-González A, TorresNiño J (2012) AKNOBAS: a knowledge-based segmentation recommender system based on intelligent data mining techniques. Comput Sci Inf Syst 9(2):713–740 2. Han J, Kamber M (2006) Data mining concepts and techniques, 2nd edn, pp 227–378. Morgan Kaufman 3. Agrawal R, Malinski T, Swami A (1993) Mining associations between sets of items in large databases. In: Proceedings of the ACM SIGMOD international conference on management of data, pp 207–216 4. Teimoriasl HS (2007) Tehran stock exchange index prediction using artificial neural networks. Iran Account Auditing Reviewing 1024–8161 5. SuryaNarayana G, Kolli K, Ansari MD, Gunjan VK (2021) A traditional analysis for efficient data mining with integrated association mining into regression techniques. In: ICCCE 2020, pp 1393–1404. Springer, Singapore 6. Arezoo A, Ali S (2009) Using Bayesian networks for bankruptcy prediction: empirical evidence from Iranian Companies. ISBN: 978-0-7695-3595-1 7. Kottapalle P, Maddala S, Gunjan VK (2016) D-mine: accurate discovery of large pattern sequences from biological datasets. In: Proceedings of the international conference on soft computing systems, pp 647–661. Springer, New Delhi 8. Agrawal and Schor ling (1996), West et al (1997), Aiken and Bsat (1999), Wang (1999), Jiang et al (2000), Vellido et al (1999) Selected customers, market share, packaging market and market trends 9. Tiwari MS, Murthy VMSR, Raina AK (2020) Challenges in mining industry and addressing through research and innovation. Helix 10(01):38–42 10. Wang and Leu (1996), Wittenmyer and Steiner (1996), Desai and Bharati (1998), Saad et al (1998), Qi (1999), Leung et al (2000b), Chen et al (2003); Predicted direction, Index, returns, risk, Rate of change, Futures stocks and commodity prices, e ciency and ... Stock market 11. Pourhassan A, Minaei-Bidgoli (2007) 3, Classification using the combined association rules and classification. Amir Kabir University 12. Quinlan JR (1992) C4.5: Program for machine learning. CA. Morgan Kaufmann, San Francisco 13. Kant R, Karmore JS (2020) Qualitative Analysis of Groundwater from Nandgaon Peth Village District Amravati, Maharashtra. Helix 10(01):56–59 14. Clark P, Niblitt T (1989) The CN2 induction algorithm. Mach Learn 3(4):261–283 15. Cohen W (1995) Fast effective rule induction. In: Proceedings twelfth international conference on machine learning, pp 115–123

A Review on Stock Market Analysis Using Association Rule Mining

183

16. Duda R, Hart P (1973) Pattern classification and scene analysis. Wiley, New York 17. Uttarwar MD, Devendar M (2020) Investigation into the effect of stemming on blast performance in underground excavations-a model study. Helix 10(01):43–50 18. Singh HK, Geete SS, Tiwari MS, Rajurkar VJ (2020) Review of limit equilibrium methods for stability analysis of dump slope. Helix 10(01):89–97 19. Rani P, Mishra AR, Ansari MD, Ali J (2021) Assessment of performance of telecom service providers using intuitionistic fuzzy grey relational analysis framework (IF-GRA). Soft Comput 25(3):1983–1993

Student Performance Assessment Using AI K. L. S. Soujanya, Challa MadhaviLatha, M. Swathi, Ch. Mallikarjuna Rao, and Sridevi Sakhamuri

1 Introduction Since 2006, the deep learning (DL) has promising theme in machine learning (ML) [1]. DL is an eminent than other algorithms of ML due to various reasons such as capacity to handle huge data, auto-learn feature, and do processing with ordered and unordered data [2]. As indicated by Deng and Yu [1], DL is incredible as far as forecast, order, recognizable proof, and recognition. As far as forecast, DL has been actualized in numerous zones to illuminate a few issues, for example, in the intelligent transport system, just as different regions of economy and training [3]. Other than expectation, DL is likewise ground-breaking for ordering, recognizing, and distinguishing an enormous number of information. Be that as it may, those four elements of DL can be utilized together and conversely in a solitary report, for example, grouping, recognizing, or distinguishing information before anticipating what will occur later on [4]. In the instruction field, a few investigations have utilized DL to defeat issues in learning and advanced education (Xing and Du 2018; Patil et al. 2017; Ojha and Heileman 2017; Wang et al. 2017) [5–8]. There are a ton of issues in training, particularly advanced education, which should be tended to. Advanced learning has a job to play in improving the quality of HR by providing graduates with abilities and capacities that are expected to be used to contribute to the country’s turn of events. Be that as it may, advanced education faces normal issues in instructing understudies who may K. L. S. Soujanya (B) · C. MadhaviLatha · M. Swathi CMR College of Engineering & Technology, Kandlakoya, Hyderabad, India e-mail: [email protected] Ch. Mallikarjuna Rao CSE, Gokaraju Rangaraju Institute of Engineering and Technology, Hyderabad, India S. Sakhamuri Department of Electronics and Computer Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, Guntur District, Andhra Pradesh, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_21

185

186

K. L. S. Soujanya et al.

neglect to complete their examinations or may drop out (Graham 2017; Weale 2018) [9, 10], understudies who may concentrate inside a more drawn out time than typical, and understudies who complete their investigations however without adequate abilities to contend. For this situation, the scholastic execution of understudies is one of the standards for effective teaching in universities [11]. One more issue in training is absence of set abilities. In an educational plan, there are aptitudes that are conveyed to various courses [12, 13]. In the wake of finishing their examinations, understudies are relied upon to have those aptitudes, for example, diagnostic abilities, the board aptitudes, or programming abilities. In any case, in the wake of graduating, once in a while those aptitudes are not adequate. Also, there may be a hole between abilities set in the educational program and aptitudes required in the business [1, 14] because of the innovation advancement. This marvel presents huge difficulties to further extent organizations, and they need to investigate the best approach to deal with those problems and different variables that influence the success rate of graduates, just as to anticipate effective alumni dependent on their practices while learning at a college [18]. Presently, grade point average (GPA) is utilized as a norm of understudies’ prosperity. Each further extent establishment has its own arrangement with respect to the base GPA that understudies should pick up to succeed in all subjects [10, 15, 16]. In the primer report, this examination utilizes log information from a course to foresee whether understudies get a great score, get a middle-of-the-road score, or fall flat. Later on, it is normal that DL can be utilized for anticipating performance of pupils’ education (PPE). The association of this paper is as per the following. Segment 2 talks about a few related works with respect to deep learning calculation for expectation reason. Segment 3 examines research techniques, dataset, and pre-processing. Area 4 talks about the consequences of our trials. At long last, the last segment gives the end and proposals to future exploration.

2 Literature Survey According to Patterson and Gibson [2], DL is defined as a neural system with more neurons, more shrouded layers, and more associated layers working as a calculation that trains and learns naturally. Deep learning is also known as the deep neural network. In any case, there is no clear definition of how many layers should be used at this time. Figure 1 depicts deep learning engineering with six highlights as information sources, n hidden layers with n hidden units in each concealed layer, and one yield layer. Profound studying calculations are ordered into three sorts: managed, solo, and half and half [1]. In their application, DL calculations are utilized for different purposes, to be specific forecast, recognizable proof, identification, and grouping. A few examinations utilizing the DL calculation that come nearer from different

Student Performance Assessment Using AI

187

Fig. 1 Architecture

fields were checked on to search for models that could fill in as a motivation for this investigation. Distinguishing proof uses solo preparing calculations, while recognition utilizes administered preparing calculations. As per Patterson and Gibson [2], DL is characterized as a neural system with more neurons, more shrouded layers, and more associated layers working as a computational calculation that trains and learns includes consequently. DL is additionally called the deep neural network. Notwithstanding, as of recently there is no express meaning of what number of layers ought to be utilized in DL. A few examinations utilizing the DL calculation that come nearer from different fields were investigated to search for models that could fill in as a motivation for this investigation. Recognizable proof uses solo preparing calculations, while identification utilizes regulated preparing calculations.

3 Models Inspected utilizing an exploratory examination strategy, the primary thought of this investigation is shown in Table 1. The forecast measurement is normal yield, which would be used as a logical educational device for getting knowledge in individual and educational plan progress so as to limit understudies’ disappointment. Different kinds of information were utilized to foresee understudies’ scholastic accomplishment, for example, individual foundation, including sexual orientation (Madhavi and Soujanya 2020) [16], race/ethnicity (Xing and Du 2018) [8], demography (Madhavi Table 1 Exploratory variable information S.no

Data_Type

Narration of data

1

Ordered

Demographic information like individual foundation, for example, sexual orientation, age, and the local location. Scholarly information, a scholarly expected test, pre-school scores, and the GPA. Social information

2

Unordered

LMS for individual intercession, the yield of this framework incorporates the likelihood for every pupil

188

K. L. S. Soujanya et al.

and Soujanya 2021) [15], earlier scholarly examination (Dong et al. 2018; Lu et. al., 2014) [12], and log information (Xing and Du 2018; Lu et al. 2014; Santoso and Putra 2017) [8, 14, 17]. Notwithstanding, in this exploration, we distinguished a few variables to be examined as recorded in Table 1. Demographic information incorporates individual foundation, for example, sexual orientation, age, and the local location. Scholarly information means all information identified with understudies’ scholastic exercises, for example, a scholarly expected test, pre-school scores, and the GPA of the understudies acquired while graduating. At that point, social information incorporates log information from LMS and recorded information from the library’s data framework [3]. Unstructured information was recovered from a conversation discussion on LMS. Finally, information from graduated class was recovered to dissect their practices after graduation. As clarified in the past area, for example that for additional exploration this model will be actualized in LMS for individual intercession, the yield of this framework incorporates the likelihood for every understudy [13] to come up short, to complete his/her examinations, to about come up short, and to graduate with absence of set aptitudes. Data collection. Crude information is taken from Software Engineering, Department of Computer Science, Universitas Indonesia. The talk was conveyed utilizing a double mode approach, to be specific a mix up of close and personal and online meetings through LMS. Information was basically gathered from the log information on LMS and scholastic scores *). Log information was utilized. Several kinds of action can be recovered from log information, for example, regardless of whether they post in discussion or not, whether they read the gathering, whether they see prospectus, get to sound, how often they endeavor to do online test, and how regularly they study. These exercises show understudies’ practices and learning movement and may be elements of understudies’ scholastic accomplishment. As can be seen, log data from week one to week four is used to forecast the seventh week. Dataset collection Approach. Generally speaking, in this examination, crude information is changed into a dataset. Information pre-handling, information purging, exploratory data analysis (EDA), include designing and different strategies are directed. (1)

Data pre-processing Log information is assembled dependent on the kind of action in LMS. Missing worth, conflicting information, clamor, and exception are analyzed in this progression. The way toward totaling information is additionally remembered for this progression. In this investigation, it is discovered that five understudies are in scholastic leave; along these lines, the last score is under 65; and however, their log information is dynamic. Another issue, log information contains the absolute hits of any movement. First endeavor, dataset is manufactured dependent on the complete hits of action to be included as appeared in Fig. 2. Subsequent to building the dataset, it is elusive the example. Along these lines, more

Student Performance Assessment Using AI

189

Fig. 2 Activity chart

(2)

top-to-bottom investigation is led. A few exercises from log information are gathered and dissected every week. Analyzing Exploratory Data (AED)

(3)

AED is basic to look at all factors. This progression permits to imagine information and show numerical rundown in measurements, for example, mean, middle, etc. The outcomes and ramifications of probing information investigation are clarified in the following segment. Technical Features Highlight building permits to choose, connect, and look at which significant highlights for model. Much of the time, include choice is expected to build the presentation of forecast. Table 2 shows the rundown of highlights.

4 Investigation At the past investigation, we constructed dataset dependent on the all out hits of understudies in LMS. Be that as it may, the complete mouse snap of understudies in log information could not reflect intellectual exercises. This is in accordance with

190

K. L. S. Soujanya et al.

Table 2 Specifications of the dataset Feature

Specification

Data_Type Particulars

Gender

Pupil gender

Nr

Level_discussion

The level of discussion for less than the average Or of student’s performance

H/M/L

W1_Discussion

Week 1 discussion about performance

Or

1 0

W2_Discussion

Week 2 discussion about performance

Or

1 0

W3_Discussion

Week 3 discussion about performance

Or

1 0

W4_Discussion

Week 41 discussion about performance

Or

1 0

Or

1 0

Read_Discussionw1 Week 1 read forum discussion

F/M

Note Nr—Nominal, Or—Ordinal, H—High, M—Medium, L—Low, 1: Yes, 0: No

a past report from Mayr et al. 2016 [4]. Be that as it may, from this examination, a few focuses can be deciphered to mixed learning. Initially, it is very hard to foresee the understudies’ last score dependent on log information, particularly if just think about the all out hit of any movement. Many factors should be considered, such as time frame for study by the LMS, whether they control their self-study, whether they submit tests ahead of time or not, prior information, and so on. Many factors should be considered, for example, how long they study using LMS, whether they control their self-study, whether they submit test prior or not, previous information, and so on. This investigation utilizes a few elements to manufacture the dataset as recorded in Table 2. The highlights are investigated by observing the connection. It very well may be seen that each element has no relationship, low connection, and somewhat high relationship to another component. Generally, as appeared in Fig. 4, all highlights appear to have feeble connection to the yield. It is faulty whether the highlights picked can speak to the class of conclusive score or not (Fig. 3). Another perception is referring the total review modules, which will be shown as how frequently the student considers, likewise shifts among understudies who scored low grade and high grade. However, as shown in Fig. 5, understudies with lower scores appear to have more review in course than understudies with higher scores. In this situation, they may contain a superior ability to investigate and a high level of curiosity than high-scoring understudies. Results: See Figs. 2, 3, 4, and 5.

Student Performance Assessment Using AI

191

Fig. 3 Score

Fig. 4 SVM accuracy

5 Conclusion When all is said in done, there are expected advantages from this examination. Right off the bat, this examination manufactured a course-level dataset for PPE; second, while this paper led a primer report utilizing DL to anticipate PPE, the outcomes are still low precision on the grounds that the amount of information is most likely insufficient for DL and highlights utilized in this examination have not spoken to PPE. As per probing information investigation, this investigation shows that log information would not be reflecting psychological exercises, particularly if the pace of highlight use was just determined dependent on the complete hits of each movement on LMS. Be that as it may, a few bits of knowledge can be increased, for example, the normal and average scored student will get high inspiration to concentrate than the high score people. It is suggested that for further investigation, a few aspects would reach higher correlation with significant scores. Utilizing bigger information from understudies

192

K. L. S. Soujanya et al.

Fig. 5 CNN accuracy

utilizing DL calculation is being suggested, for example, social, segment, and monetary components. It is likewise proposed to manufacture a prescient model that remembers other information as recorded for Table 2.

References 1. Deng L, Yu D (2014) Profound learning: methods and applications. In: Foundations and trends® in signal processing, pp 197–387 2. Patterson J, Gibson A (2017) Deep learning: a practitioner’s approach. O’Reilly Media Inc. 3. Xing W, Du D (2018) Dropout prediction in MOOCs: using deep learning for personalized intervention. J Edu Comput Res 4. Mayr A, Klambauer G, Unterthiner T, Hochreiter S (2016) DeepTox: toxicity prediction utilizing deep learning, 3(February):1–15 5. Patil P, Ganesan K, Kanavalli A (2017)Powerful deep learning model to predict student grade point averages. In: 2017 IEEE international conference on computational intelligence and computing research (ICCIC), 2017, pp 1–6 6. Weale S (2018) College drop-out rates in UK ascend for third progressive year, 8 Mar 2018 [Online]. Accessible: https://www.theguardian.com/training/2018/damage/08/col legedrop-out-rates-uk-rise-third-year. Accessed 22 Jan 2019

Student Performance Assessment Using AI

193

7. Paul PK, Solanki VK, Gunjan VK (2020) Need of iSchools in developing countries 8. Ojha T, Heileman GL, Martinez-Ramon M, Thin A (2017)Forecast of graduation postpone dependent on understudy execution. In: 2017 International joint conference on neural networks (IJCNN), pp 3454–3460 9. Graham (2017) Age dropout: record low for unifulfillment, 30 Nov 2017 [Online]. Accessible: https://www.news.com.au/wayoflife/reality/news-life/aussie-understudiesleaving-colleg ein-record-numbers/report/3d9c6b0488174b8b4107f05ce03fd6a0. Accessed: 22 Jan 2019 10. Blentsov IV, Zheleznikova OE, Voynova OS, Batarshev DS (2020) Optimization of lighting conditions with LED light sources. Helix 10(02):87–93 11. Lv Y, Duan Y, Kang W, Li Z, Wang F (2014) Traffic flow prediction with big data: a deep learning approach, pp 1–9 12. Dong, Shao C, Clarke DB, Nambisan SS (2018) An imaginative methodology for car accident estimation and expectation on obliging surreptitiously heterogeneities. Transp Res Part B Methodol 118:407–428 13. Rodrigues F, Markou I, Pereira FC (2019) Joining time-arrangement and printed information for taxi request expectation in occasion zones: a profound learning approach. Inf Combination 49:120–129 14. Santoso HB, Putra POH (2017) Overcoming any barrier between IT graduate profiles and job requirements: a work in progress. In: 2017 seventh world engineering education discussion, pp 145–148 15. Lukoyanova MA (2019) Formation of ICT competency of bachelor students while studying the course “Information Technologies” in education. Helix 9(04):5182–5186 16. Ivchenko OA, Pankin KE, Kusmartseva EV, Anisimov SA, Tutin AV (2020) Experimental studies on the flameproofing efficiency of some inorganic substances upon the inflammation of wood and fibrous materials. Helix 10(05):109–113 17. Wang W, Yu H, Miao C (2017) Profound model for dropout prediction in MOOCs. In: Proceedings of the 2nd international conference on crowd science and engineering, 2017, pp 26–32

Automatic Smoke Absorber and Filter Venu Adepu, V. Ranga Sai Kiriti, K. Veera Bhadra, N. Sai Deepak, and P. S. G. Aruna Sri

1 Introduction We often assume that an open door or a running fan would suffice to allow smoke into the area; however, although a running fan will help disperse air inside the room, it may not ventilate the entire space [1]. When considering ventilation products such as exhaust fans, they would be mounted at the top of the building, allowing smoke to be removed only if it rises to the top and if the smoke is sent out, the air pressure outside the room (outer environment) is higher than the air pressure within the room, so the air (smoke) is sent out, implying that there is still a small amount of smoke present in the air inside the room [2]. As a result, open windows or fans do not effectively ventilate the room. To minimize the risk of contaminated air and unsafe smoke in living conditions such as homes and workplaces, as well as work environments such as offices and every other enclosed space, we need an effective smoke absorber that forces out all of the smoke and air and filters it, resulting in clean and protected air within the room. Many new jobs and goods, as well as factories and other facilities, are produced as a result of rapid technological advancement; therefore, this invention, which is meant to make human life better, does not hurt it. We should use technologies to keep us safe and secure, at least in our homes and workplaces, from risks such as breathing dirty air and smoke. As a result, a smoke absorber with an integrated fume extraction fan and a carbon filter is required to efficiently emit smoke out of the room.

V. Adepu (B) · V. Ranga Sai Kiriti · K. Veera Bhadra · N. Sai Deepak CMR College of Engineering and Technology, Kandlakoya, Hyderabad, India e-mail: [email protected] P. S. G. Aruna Sri Koneru Lakshmaiah Education Foundation, Vaddeswaram, Guntur District, Andhra Pradesh, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_22

195

196

V. Adepu et al.

2 Literature Review When wood and other organic materials fired, smoke is created which is made up of a complex mixture of gases and fine particles. Small particles are the most dangerous component of smoking for your lungs. These small particles have the ability to penetrate deep into the lungs. They can trigger anything from burning eyes and a runny nose to worsening chronic heart and lung disorders [3]. It is critical to keep your exposure to smoke to a minimum, particularly if you are susceptible to particle-related health problems. While very thick smoke cannot be purified from oxygen, it can be transported from within a room to the outside world, preventing people from inhaling the toxic smoke and causing breathlessness and other respiratory problems. The primary purpose of a smoke absorber is to trap smoke from the air inside, so filtration is secondary; however, filtration can only be achieved to a certain degree with smoke, and thick smoke cannot be filtered. Both air purifiers use active carbon filter sheets to collect small particles from the air and give out clear air [4], while exhaust fans use rotating fans to trap smoke from the room and emit it from the other side [2]. Integrating these two functions and creating a device that can fulfill both functions effectively bring us many benefits.

2.1 Problem Statement Since the majority of people live in countries with low air conditions, we know that clear and clean air is essential for a safe human being and one of the most important requirements for any human community.

2.2 Objectives The key goal is to develop a substance that can trap smoke from a space (closed environment), provide clean air to breathe within the room, and purify the smoke before releasing it into the outside world. To clear smoke, grease, or gases from a room, thereby shielding people from unhealthy smoke and reducing the risk of becoming sick from airborne diseases [15].

Automatic Smoke Absorber and Filter

197

2.3 Existing Solutions 2.3.1

Air Purifiers

An air purifier, also known as an air cleaner, is a system that eliminates pollutants from the air in a room to increase indoor air quality. These devices are often sold as being useful to allergy and asthma sufferers, as well as for limiting or removing second-hand cigarette smoke. Commercially graded air purifiers are available as small stand-alone units or larger units that can be attached to an air handler unit (AHU) or a heating, ventilation, and air conditioning (HVAC) unit used in the medical, manufacturing, and commercial sectors [5, 6]. Air purifiers may also be used in industry to eliminate impurities from air prior to processing. Pressure swing absorbers or other adsorption techniques are often used. By size exclusion, air filter purification captures airborne particles. Air is pushed through a filter, which physically captures particulates [7, 17]. There are several filters available, most notably: HEPA filters remove at least 99.97% of 0.3-µm particles and are generally more efficient at eliminating bigger particles [8, 16]. HEPA air purifiers, which filter the entire atmosphere, going into a clean room must be set up in such a way that no air bypasses the HEPA filter. In dusty settings, a HEPA filter may be used after an easily cleaned conventional filter (pre-filter) that eliminates coarser particles, reducing the frequency with which the HEPA filter needs to be cleaned or replaced. HEPA filters do not produce ozone or other hazardous byproducts during operation. Activated carbon is a porous material. When employing activated carbon, the adsorption process must attain equilibrium, making it difficult to entirely remove pollutants [9, 18]. Activated carbon is simply a method of converting pollutants from a gaseous to a solid state; contaminants can be regenerated in indoor air sources when they are exacerbated or disturbed. Activated carbon has a long history of commercial application and may be utilized at room temperature. It is typically used in combination with other filter technologies, particularly HEPA. Other materials can absorb chemicals as well, but at a higher cost.

2.3.2

Exhaust Fans

The use of range-exhaust fans while cooking will help improve indoor air quality if the fans are vented to the outside of the house. None is known about the real prevalence of kitchen fans [10, 11]. It was discovered that one-third of the houses had no range-exhaust fans and the other half have vented fans. Fewer than half of families with fans use them on a daily basis, and even fewer use them while cooking begins. Experiments at the research facility revealed that a vented hood fan may lower peak concentrations of combustion products by approximately 50%, given that it was switched on at the start of cooking episodes. However, only approximately 12% of the studied homes

198

V. Adepu et al.

benefit from having vented fans, utilizing them consistently, and turning them on as soon as cooking begins. Reparable particulate and carbon monoxide levels were statistically higher in households with apparent leaks or back-drafting stoves. Interior NO2 levels were statistically greater during wood burning [12, 13]; however, this was attributed to increase outside levels rather than direct indoor emissions.

3 Project Design 3.1 Requirement Analysis • • • • • • • •

DC motors Fans Connecting wires Cardboard compartment Air filter Switches Batteries Hose (Fig. 1)

Fig. 1 Components

Automatic Smoke Absorber and Filter

199

3.2 Implementation The product’s basic approach is plain absorption and release of air. The cardboard compartment must be separated into two halves. An air filter separates these two halves. Two holes are drilled into the face of the container’s first half, and another hole is drilled into the face of the compartment in the second half. Absorption is carried out using two DC motors to which plastic fans are glued and mounted so that they spin along with the motor when the motors are provided power supply; these two DC motors are arranged in such a manner that they are away from the face of the cardboard box from the inside when the motors are given power supply. When the DC motor is driven by batteries, the fans spin, allowing air from outside to reach the compartment and fill it. There is a filter between the two halves of the compartment, and the other half has another DC motor connected with a fan. This would extract the air (smoke) from the first half through the filter and blow the air from the second half outside through a nozzle. The compartment is completely made of recycled cardboard and can be easily replaced with any other material. First, the cardboard compartment is divided into two halves and measurements are taken so that the air is distributed among the two halves equally. Two holes are cut on the first half of the compartment such that the diameter of the hole is more than the diameter of the fan; this way the fans can easily suck in the air without any difficulty. One hole is cut on the face of the second half of the compartment such that the diameter of this hole is greater than the diameter of the fan so that the fan can easily emit the air outside. A 13 × 10 cm rectangular space is left in between the two halves of the compartment; the remaining portion is covered with cardboard. This compartment is made such that it air tight inside, and no air leaks outside except through the hole on the second half of the compartment. The circuit is made up of three 3 DC motors, a switch, and five 9 V batteries. The first two DC motors are linked in parallel so that they rotate at the same speed, and three 9 V batteries are connected in series and connected to the switch so that they transfer electricity when the switch is turned on. When the switch is turned on, the three 9 V batteries linked in series are connected to the two DC motors connected in parallel, causing them to revolve. A DC motor is installed in the second half of the container, and two 9 V batteries are linked in series. These batteries are connected to the DC motor, which is attached to a switch and begins rotating when the switch is turned on. All of the connections are made via connecting wires. The DC motors are maintained in front of the holes produced in the first half of the compartment; the DC motor is spaced such that the fans face away from the holes made. These two are stored next to each other, and the three batteries linked to them are put at one end of the cardboard box, outside, so that they may be withdrawn and attached to the battery caps when totally drained. The remaining DC motor is linked to the fan and is positioned such that the fan faces the hole on the face from the inside of the compartment.

200

V. Adepu et al.

Fig. 2 Prototype of smoke absorber

This is linked to the switch, and the batteries are connected outside the cardboard box so that they may be replaced when they are discharged. When the switch is turned on, all of the motors begin to revolve at the same time (Fig. 2).

3.3 Block Diagram See Fig. 3.

3.4 Applications 3.4.1 1. 2. 3. 4. 5.

Air Filter

Air filters are used in gas/air flow systems to separate particulates from the air flow, which is referred to as “air conditioning.” Smoke absorbers serve a number of functions, some of which are described below. Enhancing the user’s breathing condition. Defending against airborne particle damage to walls, ceilings, and machinery, among other things. Smoke absorbers are used to remove smoke, foul odors, hot air, and other poisonous air from a closed space.

Automatic Smoke Absorber and Filter

201

Fig. 3 Block diagram of smoke absorber

3.4.2 1.

2.

3.

Exhausting the Air

A smoke absorber may be used to draw excess moisture and odors out of a specific space or location. They are commonly found in bathrooms and kitchens, where moisture can accumulate. A smoke absorber can easily cool off areas that have been overheated as a result of things such as cooking or showering. Hot air is vented outside, lowering the temperature of the room without using the air conditioning system. Excess moisture that can harm the home can be removed by using a smoke absorber. Condensation from hot water use may accumulate on walls, ceilings, and other surfaces, allowing mold to emerge.

202

V. Adepu et al.

4 Conclusion With more developments, this low-cost, low-maintenance smoke absorber could replace traditional exhaust fans in many areas. It is also a highly versatile device that can totally absorb smoke from a room and emit it outside. Not only can this be used as an exhaust, but it can also be used as a filter. One of the most significant benefits of this device is that it is extremely portable; it can be used anywhere, regardless of the environment, and it can also be used outdoors to divert smoke in small areas such as campfires. As a result, it is a blend of two valuable items at a reduced cost and with less upkeep. This device is far less expensive than the exhaust fans or smoke absorbers currently available on the market.

References 1. 2. 3. 4. 5.

6. 7.

8.

9.

10.

11. 12.

13.

14.

https://blog.sra-solder.com/knowledge-base/smoke-absorber-benefits https://www.brooklynfan.com/five-common-uses-for-exhaust-fans/ https://www.munters.com/en/munters/case-studies/schering-ag-pharmaceutical/ https://cowaymega.com/blogs/blog/what-is-a-hepa-filter Noussan M, Carioni G, Degiorgis L, Jarre M, Tronville P (2017) Operational performance of an air handling unit: insights from a data analysis. Energy Proc 134:386–393. https://doi.org/ 10.1016/j.egypro.2017.09.579 Qian H, Li Y, Sun H (2010) Particle removal efficiency of the portable HEPA air cleaner in a simulated hospital ward. Build Simul 3:215–224. https://doi.org/10.1007/s12273-010-0005-4 Liu G, Xiao M, Zhang X, Gal C, Chen X, Liu L, Pan S, Wu J, Tang L, Clements-Croome D (2017) A review of air filtration technologies for sustainable and healthy building ventilation. Sustain Cities Soc 32. https://doi.org/10.1016/j.scs.2017.04.011 Cerro G, Ferdinandi M, Ferrigno L, Laracca M, Molinara M (2018) Metrological characterization of a novel microsensor platform for activated carbon filters monitoring. IEEE Trans Instrum Meas 67(10):2504–2515. https://doi.org/10.1109/TIM.2018.2843218 Lee JH, Kim JY, Cho BB (2019) Assessment of air purifier on efficient removal of airborne bacteria, Staphylococcus epidermidis, using single-chamber method. Environ Monit Assess 191:720. https://doi.org/10.1007/s10661-019-7876-3 Silva W, Moura D, Carvalho-Curi T, Seber R, Massari J (2017) Evatuation system of exhaust fans used on ventilation system in commercial broiler house. EngenhariaAgrícola 37:887–899. https://doi.org/10.1590/1809-4430-eng.agric.v37n5p887-899/2017 Patnaik A, Ali SM (2013) Industrial exhaust fans as source of power. Int J Electr Electron Data Commun (IJEEDC) 1(9):38–41 Iyer US, Raj PE (2013) Ventilation coefficient trends in the recent decades over four major Indian metropolitan cities. J Earth Syst Sci 122:537–549. https://doi.org/10.1007/s12040-0130270-6 Curi TMRC, Vercellino RDA, Massari JM, Souza ZM, Moura DJ (2014) Geostatistic to evaluete the environmental control in different ventilation systems in broiler houses. EngenhariaAgrícola 34(6):1062–1074 https://en.wikipedia.org/wiki/Compressed_air_filters#Working_principle

Computational Intelligence in Subthalamic Nucleus Deep Brain Stimulation: Machine Learning Unsupervised PCA Tracking Method and Clustering Techniques for Parkinson’s Feature Extraction Venkateshwarla Rama Raju, B. Anuradha, and B. Sreenivas

1 Introduction Parkinson’s disease (PD) is a progressive chronic neurodegenerative syndrome typified by cardinal motor signs: tremor, rigidity, akinesia (bradykinesia) plus postural instability. Since, no definitive test is existing for curing Parkinson’s, and the clinical diagnosis is derived from and rooted as of the occurrence of clinical signs plus retort to anti-Parkinsonian medicine [1]. The majority of the conventional scale for evaluating disability and injury in PD is the Unified Parkinson’s Disease Rating Scale (UPDRS) stage III+ [2] and is rooted and founded on biased medical estimation of syndromes. Hence, it necessitates quantifying PD characteristics scientifically in line, in sequence to advance “diagnosis”, define “disease subtypes”, observe this malady evolution or succession, and also exhibit behavioral management effectiveness [3, 4]. DBS is a neurosurgical process for Parkinson’s remedy which employs highfrequency square pulses to excite STN neurons plus connected substructures and subregions of the human brain. Even though the means and methods of the deep brain simulator act are not comprehensive, exact neurochip implantation and stimulus coding and encoding might progress motor signs and also let for a decrease in anti-Parkinsonian dosage medicine [5]. But, the stimulus parameters are positioned by conventional guess-of-signs (GoS), and also, no physiological capacitative magnitude gages are applied to maximize the value of stimulator for reducing motor syndromes [6]. Non-invasive EMG plus kinematic gages facilitate the scientific computation of neuromuscular functioning, and action and movement progress, therefore, can be V. R. Raju (B) · B. Anuradha · B. Sreenivas Department of Computer Science and Engineering, CMR College of Engineering & Technology, Medchal Rd.501 401, Kandlakoya, Hyderabad, Telangana, India e-mail: [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_23

203

204

V. R. Raju et al.

applied for computing the outcomes of deep stimulator of the brain, anti-Parkinsonian medicine, else supplementary curing. Earlier experiments have shown that myography distinctiveness of subjects by this malady might transform because of owing to stimulator and anti-Parkinsonian medicine in as a minimum of three modes. Initially, the principal shaking frequency in the myogram range, i.e., frequency range is augmented by stimulator of the brain followed by medicine [7, 8]. The, then, coherence amid electromyography acceleration signifying tremor is condensed by the stimulator, drug in latent state and also outward [8, 9]. Lastly, period, stimulus amplitude pulse width of the primary distress disintegrate is amplified, then the numerous distress fragments is compacted by the brain stimulator plus medicine through quick point-to-point movement actions of jostle/prod [10] followed with ankle [11]. On the other hand, yet the groupings of stimulator and the drug cannot standardize the electromyography fracture singularity [11]. Also, previous experimental investigations have demonstrated that nonlinear and morphological techniques of myography plus acceleration inferences in convolution through PCA approach (via execution of K-L transform) are greatly helpful for discerning subjects by means of in the middle of the maladies of normal controls [12– 15]. This consequence is of the reality that the myograms of subjects by Parkinson’s vary as of the myographs of normal controls, presetting spiny, inveterate anatomical structures followed by the acceleration gatherings measurements depicting reliability. But, this is not weathered whether approximating techniques of nonlinear dynamics followed by the K-L implemented latent variate-based (PC) approach are competent for computing outcomes of stimulator scientifically. In this paper (experimental investigative research study), we demonstrate machine learning-based unsupervised PCA and cluster tracking method for computing the outcome of deep brain stimulator in Parkinson’s by means of applying myogram followed by kinematic investigation. Ten parameters capturing Parkinson’s movement disorder distinctive signal feature-manifestations were originally extrapolated as of isometric myography plus quickening signal recordings, and those features (signals) were parameters of nonlinear dynamics, coherence amid myography, quickening plus stimulus amplitude pulse widths of increase of velocity (the acceleration). By employing the PC approach, the original parameters were transformed into a lesser utmost at best four principal components. However, we considered first three PCs only. At last, the results of brain stimulator were computed by observing the P C components in a low-dimensional feature space (FS). The myography plus increase of velocity data as of 12 Parkinson’s in the middle of STIMULI ON and STIMULIOFF plus 12 normal healthy controls were employed for deducing the inferences. The outcomes of STIMULI on jostle (elbow) flexion also on (extension) porchactional movements were inferred independently. The following table shows the clinical characteristics of the PD subjects.

Computational Intelligence in Subthalamic Nucleus …

205

2 Hypothesis Deep brain stimulation implants the micro- and neurosensors or intelligent chips (i.e., microelectrodes) into an important area of the brain “substantia nigra (SN)”. The STN neurons are surrounded by SN. So, whether the implanted pulse generators damage the STN neurons or save and reduce the motor symptoms. Hence, the hypothesis of this study is that the built-up PC method is capable of computing the outcomes of DB stimulator on subjects with PD objectively. It follows that the measurement characteristics of Parkinson’s are highly akin to measurement characteristics of normal controls (healthy) with “BRAIN STIMULATOR ON” than with “BRAIN STIMULATOR OFF”.

3 Methods The following machine learning unsupervised computational techniques are applied in this study.

3.1 Latent Variate Factorial PCs The microsignal, i.e., MER signal features of subthalamic-nuclei (STN) neurons of the 12 Parkinson subjects showed in Fig. 1 and their clinical characteristics demographics in corresponding Table 1. We have implemented the Kerhnen Levin KL transform in MATLAB mainly for transforming the original correlated changeable variables into unconnected/correlated changeable variables and for dimensionality reduction to decrease the number of changeables while observance as much as possible mostly informative concerning original variables. Upon computing, firstly ten signal parameters were positioned in sequence plus normalized to 0 mean plus unit standard deviation (SD) of normal’s to shape FVs for every subject. Individual FV was shaped for every normal with two FVs for every subject: one by “stimulator on”, one by “stimulator off”. Then, the dimensional measurement of the FVs was decreased through PC approach. Because in such given move toward advance, the FVs were singularly decomposed (singular value decomposition) into biased sum of orthogonal basis vectors, where ever the scalar weights were referred to as the dominant PCs and they were the new and also uncorrelated and unassociated features and unconnected vectors. We have chosen the basis vectors that were here by employing the FVs of normal’s, i.e., healthy people. By choosing this selection, we ranged the features of normal’s within a single standard deviation. So, we clustered the normal’s into one conglomerate for comparative future study purposes, and for computing the basis vectors, we produced a feature matrix which limited the FVs of normal’s within its column.

206

V. R. Raju et al.

Fig. 1 Myography plus IoV feature values for a single normal person depicted with *asterisk and a diseased condition with PD brain stimulator on/off depicted with circle (°)

Later, we computed the correspondence matrix which is the associated or connected, correlated of the feature matrix, and we also computed the corresponding dominant eigen vectors (EVs) as of that. Five EVs analogous to five highest magnitudes eigenvalues were selected as “basis vectors”. These five EVs were given the 96% of the overall disparity in “FVs” of every normal. Therefore, every FV possibly will be fairly precisely formed as a biased sum of these five EVs. Lastly, we computed PCs in a least square sense method for every normal and for every PD subject by “stimulator on” and “stimulator off”. The best PCs to discriminate among stimulator on, stimulator off states plus connecting normal’s, subjects of Parkinson’s were preferred to deduce the inferences later on. It is this PCs that are offered in a 2D “feature space” plus contrasted connecting “stimulator on” plus “stimulator off states” as well as among Parkinson’s and normal’s. Consistent with the hypothesis of this study, in the FS, the subjects shall be nearer to conglomerate of normal’s in the middle of “stimulator on” than through the “stimulator off”.

3.2 Principal Components PCs were computed for every PDs through the evaluated EVs. It is pointed that the PC1 plus PC3, i.e., first principal component PC1 and third principal component PC3 function and yield best in discerning among the brain stimulator on, brain stimulator off states, plus among the PD subjects, normal’s.

Computational Intelligence in Subthalamic Nucleus …

207

Table 1 Clinical data of PD subjects Movement disorders and Parkinson’s disease* Case no.

Age (yrs.)

Sex

Occupation

Disease

Duration (yrs.)

Date of recording

Organ (neural)

PDS1

71

Male

Prof (neurosurgery)

PD

12

16/02/15

Bilateral STN

PDS2

61

Male

Medical doctor

PD

5

08/02/15

Bilateral STN

PDS3

63

Male

Physician

PD

3

15/06/14

Bilateral STN

PDS4

60

Male

Bank employee

PD

6

29/01/16

Bilateral STN

PDS5

65

Male

Politician

PD

2

03/01/18

Bilateral STN

PDS6

62

Male

Excise controller

PD

1.6

11/05/17

Bilateral STN

PDS7

85

Male

Quality controller

PD

40

03/09/15

Bilateral STN

PDS8

69

Male

Neurosurgeon

PD

5

10/08/12

Bilateral STN

PDS9

66

Male

Police officer

PD

5

15/06/12

Bilateral STN

PDS10

61

Male

Bank manager

PD

2.5

19/06/12

Bilateral STN

PDS11

69

Male

Prof (neurosurgery)

PD

10

03/05/17

Bilateral STN

PDS12

61

Male

Foreigner

PD

12

18/06/19

Bilateral STN

The primary EV is the best mean square fit (MSF) for the FVs of normal’s. Thus, PC1 (i.e., the coefficient of primary EV) expresses the stimulus amplitude pulse width of electromyography plus increase of velocity features in connection with the mean-of-normal’s (MoNs). Observing the morphology—morphological—data (through visually) of 3 eV, we may possibly distinguish that PC3 highlights the dissimilarity among the right-side (RS) and left-side (LS) changeable variables. Actually, the unilateral onset plus importunate determined irregularity (i.e., asymmetry) of signs and syndromes hold the prognostics of Parkinson’s in connection to supplementary alike diseases and disorders or syndromes [1]. PCs 3, i.e., third PCs in respect of the PC1, i.e., all first three PCs of 12 normal’s plus 12 PDs through the stimulator on, stimulator off are showed in Fig. 2. The PC values of 12 PDs are nearer to the middle-of-normal’s (MoNs), i.e., point 0,0 in the FS through the brain stimulator on than by brain stimulator off, i.e., the electromyography plus increase of velocity feature values of them find nearer to

208

V. R. Raju et al.

Fig. 2 Principal components 3. The third PC proposed with first principal components normal’s showed with *asterisk, Parkinson’s through the brain stimulator on showed with + sign, and brain stimulator off showed with circle sign. a PD subjects 4,5 plus normal’s, b. left out PD subjects and normal’s. Brain stimulator on and off states of every Parkinson’s conditions are connected with a line

Table 2 Feature values (mean ± SD) for normal’s plus Parkinson’s while brain stimulator on/off

Signal feature Controls

PD patients with PD patients with DBS on DBS off

D2 r

6.9 ± 0.8

6.1 ± 1.0

D2 ,1

6.7 ± 0.9

5.7 ± 1.4

5.4 ± 1.6 5.4 ± 2.1

%RECr

5.7 ± 3.4

9.2 ± 5.1

14.9 ± 13.6

%REC1

6.8 ± 3.9 12.7 ± 7.6

15.3 ± 14.9

RMSr · · · 103

0.4 ± 0.1

0.8 ± 0.4

2.5 ± 4.6

RMS1 · · · 103

0.4 ± 0.1

1.0 ± 0.9

6.0 ± 12.4

SampEnr

1.3 ± 0.1

0.9 ± 0.3

0.7 ± 0.3

SampEn1

1.4 ± 0.2

1.0 ± 0.4

0.8 ± 0.4

Cohr

0.6 ± 0.3

1.3 ± 0.7

1.5 ± 1.4

Coh1

0.7 ± 0.5

1.4 ± 0.8

2.2 ± 1.5

normal’s; also, the plane distinctions shrink while the “brain stimulator” is “ON” (Table 2). The expanse or detachment among the “brain stimulator on” along with “brain stimulator off” states in the FS is extremely individual. The distances to the centerof-normal’s (CoNs) are also given.

Computational Intelligence in Subthalamic Nucleus …

209

3.3 Machine Learning Unsupervised PCA and Clustering Algorithmic Techniques Principal component analysis (PCA) is a statistical (mathematics) latent variate factorial (or factor) analysis (via dynamical systems) technique (based on heuristics, i.e., trial and error-based method) mainly to deduce the latent dynamics (hidden features mainly for feature classifications purposes) from parallelly acquired higherdimensional STN neural spiking data. The PCA is a way of classifying patterns in data and expressing the data in such a way as to underline their parities and disparities. As the patterns in data can be hard to find in data of higher dimension, wherever the extra graphical grid representation is not obtainable. PCA is a powerful tool for analyzing massive data and high end; therefore, highspeed hardware is necessary to process the volume of signals (data). PCA requires that the eigenvalues and the covariance matrix be formed. The eigenvalues obtained are unique for the entire set. Indeed, it turns out that the eigen vector (EV) with the highest eigenvalue is the principle component of the dataset. The eigen vector with the largest eigenvalue is the one that will point down the middle of the data. It is the most significant relationship between the data dimensions. In general, once eigen vectors are found from the covariance matrix, the next step is to set them by eigenvalue, highest to lowest decreasing order in their magnitudes. The components with lesser significance can be ignored, as eigenvalues with small value do not result in much loss of data because they are negligible on electrical baseline (the zero line). If some of the components are left out, then the final dataset will have fewer dimensions than the original one. If originally there are n dimensions in the data, n eigenvalues and corresponding n eigen vectors are computed and then choose only the first p eigen vectors, then the final dataset has only p dimensions. Taking eigen vectors, which are not ignored, and forming a matrix with eigen vectors in the columns form a feature vector (FV). A “FV” is a vector consisting of multiple elements or features which give the characteristics of the object. EV = eig1 eig2 eig3 . . . eign

(1)

On forming the EV, get the transpose of the vector and then multiply it on the left of the original dataset, transposed. Last-data = row-EV × adjustment of row-data

(2)

where row-FV is the matrix with the eigen vectors in the columns transposed so that the eigen vectors are now in the rows, with the most significant EV at the top, and the adjustment of the row data is the mean adjusted data transposed, i.e., the data items are in each column, with each row holding a separate dimension. The principal component (PC) program computes the mean data vector from all row vectors within

210

V. R. Raju et al.

the initial data matrix supplied to the PC program. The residual data matrix was computed with the algorithm originally designed Jacobi’s method [16, 17]. Upon computing the residual data matrix, also the first three resulting principal components are accumulated on to the computer hard disk, and the variance associated with each PC is verified and accumulated in conjunction with, and the reason is to compute three principal component vectors (PCV’s ), namely PCV1 , PCV2 , and PCV3 of a class of signals—waveforms. The program arranges the data of a single signal into a matrix of the order m × n (m < n) by splitting the signal into m segments each of length n. Next, data minimization program determines PC coefficients a1 , a2 , and a3 and calls or invokes a function to operate on the following equation [12]. X = G + a1 .P1 + a2 .P2 + a3 .P3 + error

(3)

where “X” is the test phantom vector, “G” is the mean of class of phantom vectors, P1 , P2 , and P3 are the first three PCs; a1 , a2 , and a3 are the principal components coefficients, such that the following error Eq. 4 is reduced e2 =



[X ( j ) − G( j ) − a1 .P1 ( j ) − a2 .P2 ( j ) − a3 .P3 ( j )]2

(4)

4 Results and Discussion We presented a PC-based tracking method for quantifying the effects of DBS in PD by using surface EMG and acceleration measurements. The method was tested with EMG and acceleration data from 13 PD patients with DBS on and off, and 13 healthy age matched controls. Detailed analyses are discussed. The electromyography and the increase of velocity called accelerator (EMG and ACCs) dimensional measurements for the period of isometric task (IMT) for a single normal person as well as one single PD diseased subject condition through the brain stimulator on and brain stimulator off are depicted in Fig. 3. The outcomes of the isometric at high force task levels (i.e., voluntarily muscle contraction at the high force levels i.e., isometric levels) depicted that the signal characteristics of 12 PDs were highly alike to the signal characteristics of normal’s through brain stimulator on than by brain stimulator off. The distinctiveness examined in myogram, IoV signals amid PD subjects and normal’s, also among the brain stimulator on/off states recommended three issues. At the outset, myogram gauges of Parkinson’s distorted and perturbed into a highly multifaceted also restricted fewer chronic anatomical structures owing to the brain stimulator. These chronic structures are possible as a result of motoric unit harmonization in hand flexion and extensor muscles, which is trait for Parkinson’s [18]. The outcomes of the brain stimulator, i.e., deep brains stimulator have not been examined either by inferring the nonlinear

Computational Intelligence in Subthalamic Nucleus …

211

Fig. 3 Electromyography plus increase of velocity (IoV accelerator) signal recordings of a single individual Parkinson’s diseased condition through the brain stimulator off (left side: LS) plus brain stimulator on in the center also once single normal (right side: RS) for the period of the isometric— high force contraction of brachii muscle

myography features. Also, the stimulus amplitude pulse widths promptness, timekeeping, reliability, and the IoV gages measurement followed by the coherence amid electromyography plus IoV decreased because of the stimulator, and the outcomes submit to decrease in shaking palsy plus reliable through previous experiments [7–9]. Lastly, the elevation plane differentiations amid LS is followed by RS changeable,

212

V. R. Raju et al.

i.e., variables decreased through stimulator. The irregularity of signs and syndromes or feature manifestations is absolute basis for Parkinson’s [1]. The expanse among the stimulator on, stimulator off states in FS was extremely personage. Similarly, the progresses in prognostic diagnosis scores were exceedingly personage. Yet, sturdy and robust variations in the overall UPDRS stage III + cardinal motor score did not result forever in sturdy variations in the inferred PCs visa-a-vis and vice-versa. It may be because of the information that the overall UPDRS stage III + motoric score [2] is a complex score which contains of a huge number of subscores. Hence, the subscores are distinctive and divergent for dissimilar regions of the vertebra in singular actional movement circumstances. In this paper, we inferred biceps brachii (BB) muscles and arm actional movements. In proportion to the outcomes, barely some of the PDs achieve the normal’s in the FS through the stimulator on. Actually, earlier authors have proved that the inferred electromyography plus IoV features are possibly functional for discerning among Parkinson’s also normal healthy people not considering of healing through medicinal drug [13, 14]. For one subject, to whom we gaged larger stimulus amplitudes and pulse widths also high shaking palsy/tremor through the stimulator on than via stimulator off, it was obviously auxiliary as of the normal’s in the middle of stimulator on. While in clinical observations, this candidature shaking palsy/tremor score was distinct to be larger by the stimulator off than by the stimulator on which proves that IoV gage measurements are able to be dissimilar in sequence concerning the shaking palsy/tremor than the experimental eye of clinical diagnosing [19]. We have examined that the technique is perceptive to Parkinson’s through the linked and related shaking palsy/tremor. The link among the shaking palsy scores plus space to normal’s in FS is noteworthy p < 0.0012 highly significant statistically by a χ 2 ∼ = 9.2859 with a 2 degree of freedom which is highly significant at 5%. Since shaking palsy/tremor appears within 80–90% of Parkinson’s [20], so is successfully decreased by the brain stimulator deep inside [5], and so the technique we used here is well suited for Parkinson’s [21–26].

5 Conclusion As a conclusion, the voluntary contractions of the BB muscles at the isometric levels, i.e., at the high force levels of electromyography plus IoV gages measurements are highly employable in clinical sectors especially in neurocare centers for computing the outcomes of brain stimulators (deep into the brain)) on the neuromuscular function of Parkinson’s. These computations in convolution in the middle of the PC-based tracking method might be employed to calculate the outcomes of the brain stimulator objectively/scientifically, cost-effectively, and non-invasively, i.e., minimally invasive. In the future, the demonstrated approach might be tested and examined for assisting the fine-tuning of brain stimulator sets. Besides, the sensitivity of demonstrated techniques to dissimilar forms of Parkinson’s ought to be predictable further cautiously in auxiliary experimental medical studies quantifiably.

Computational Intelligence in Subthalamic Nucleus …

213

References 1. Raju VR, Mridula RK, Borgohain R (2019) Effect of microelectrode recording in accurate targeting STN with high frequency DBS in Parkinson Disease. Taylor & Francis IETE J Res 1–15. https://doi.org/10.1080/03772063 2. Sarma SV, Eden UT, Cheng ML, Williams ZM, Hu R, Eskandar E, Browne EN (2010) Using point process models to compare neural spiking activity in the subthalamic nucleus of Parkinson’s patients and a healthy primate. IEEE Trans Biomed Eng 57(6):1297–1305 3. Saxena S, Schieber MH, Thakor NV, Sarma SV (2012) Aggregate input-output models of neuronal populations. IEEE Trans Biomed Eng 59(7) 4. Brown P (2003) Oscillatory nature of human basal ganglia activity: relationship to the pathophysiology of Parkinson’s disease. Mov Disord 18(4):357363 5. Meloni G, Sen A, Abosch A, Ince NF (2015) Spatial distribution of nonlinear interactions in subthalamic nucleus local field potentials in Parkinson’s disease. In: 37th annual international conference of the IEEE engineering in medicine and biology society (EMBC), Milan, Aug 2015, pp 5557–5560 6. Huberdeau D, Walker H, Huang H, Montgomery E, Sarma SV (2011) Analysis of local field potential signals: a systems approach. In: 33rd Annual international conference of the IEEE EMBS, Boston, Massachusetts, USA, Aug 30–Sept 2011 7. Rissanen SM, Kankaanpää M, Tarvainen MP, Novak V, Novak P, Hu K, Manor B, Airaksinen O, Karjalainen PA (2011) Analysis of EMG and acceleration signals for quantifying the effects of deep brain stimulation in Parkinson’s disease. IEEE Trans Biomed Eng 58(9):1–20. https:// doi.org/10.1109/TBME.2011.2159380 8. Kostoglou K, Michmizos KP, Stathis P et al (2017) Classification and prediction of clinical improve-ment in deep brain stimulation from intra-operative microelectrode recordings. IEEE Trans Biomed Eng 64:1123–1130 9. He H, Garcia EA (2018) Learning from imbalanced data. IEEE Trans Knowl Data Eng, pp 1263–1284 10. Liu X-Y, Wu J, Zhou Z-H (2008) Exploratory under-sampling for class-imbalance learning. IEEE Trans Syst Man Cybern B. 39:539–550 11. Seiffert C, Khoshgoftaar TM, Van Hulse J, Napolitano A (2009) RUSBoost: a hybrid approach to alleviating class imbalance. IEEE Trans Syst Man Cybern A Syst Humans 40:185–197 12. Sarbaz Y, Pourakbari H (2016) A review of presented mathematical models in Parkinson’s disease: black- and gray-box models. Med Biol Eng Comput 54:855–868 13. Hans S, Kim D, Kim H, Park JW, Youn I (2016) Electrical stimulation inhibits cytosine arabinoside-induce neuronal death by preventing apoptosis in dorsal root ganglion neurons. NeuroReport 27(6):1217–1224 14. Gunjan VK, Shaik F, Venkatesh C, Amarnath M (2017) Computational methods in molecular imaging technologies. Springer, Singapore 15. Raut S, Yadav D, Paithane AN (2020) Integrated analysis of fMRI-EEG data for the detection of epilepsy disorder. Helix 10(05):21–31 16. Narayana GS, Ansari MD, Gunjan VK (2022) Instantaneous approach for evaluating the initial centers in the agricultural databases using K-means clustering algorithm. J Mobile Multimedia 43–60 17. Senatus PE, Teeple D, McClelland S (2006) A technique for minimally altering automatically based subthalamic electrode targeting by microelectrode recording. J Neurosurg, Third Edition 20:E8 18. Faraji AH, Mark Richardson R (2019) Deep brain stimulation in early parkinson disease may slow rest tremor progression. Sci Times Neurosurgery 84(1):E13–E14 19. Jancovic J (2008) Parkinson’s disease and movement disorders. Lancet Neurol 7(1):8–11 20. Grundmann K (2005) Primary torsion dystonia. Arch Neurol 62:682–685 21. Gunjan VK, Janga Reddy M, Shaik F, Hymavathi V (2018) An effective user interface image processing model for classification of Brain MRI to provide prolific healthcare. Helix J 8(3):2129–2132

214

V. R. Raju et al.

22. Nayak A, Bhardwaj J (2020) Discrete wavelet transform and bird swarm optimized bayesian multimodal medical image fusion. Helix 10(01):07–12 23. Guchi D, Edwards DR, Curry E (2007) Statistical determination of the optimal subthalamic nucleus stimulation in patients with Parkinson’s disease. J Neurosurg 106:101–110 24. Muppalaneni NB, Gunjan VK (2015) Computational intelligence techniques for comparative genomics. Springer, Singapore 25. Masimore B, Kakalios J, Redish AD (2004) Measuring fundamental frequencies in local field potentials. J Neurosci Methods 138:97–105 26. Jancovic J (2007) A textbook of movement disorders, 5th edn

Color Image Retrieval with a Weighted Adjacent Structure Model N. Koteswaramma and Y. Murali Mohan Babu

1 Introduction The usage of smart phones and digital cameras is extensively increased nowadays, and the quantity of images also increased rapidly in human life. Digital images are source of pictorial information while exploring and propagating the information. This kind of pictorial information is useful in many applications like criminal identification, geographic information system, and medical imaging [1, 2]. Nowadays, with available technology like good quality digital cameras, increase of internet speed, and reduced prices of storage devices, the amount of digital images in the databases increased surprisingly for these applications. The searching of a particular image is difficult in the database that contains huge number of digital images. For searching, it may be better to develop an image search for efficient and accurate image retrieval. To fulfill this, an efficient and automated image retrieval approaches are required. There are different image retrieval methods, among them three are discussed here: semantic-based image retrieval (SBIR), content-based image retrieval (CBIR), and text-based image retrieval (TBIR) [3]. Out of these three methods, CBIR is preferred than SBIR and TBIR of their limitations and deficiencies. TBIR is used by search engines; here, precision is low. In SBIR, the major problem is user need to tag some images which is not flexible and not adaptive. Content-based image retrieval (CBIR) accepts some low-level features like color, shape, and texture for the analysis of images. In CBIR, the retrieval precision is high. So, the searching of digital images is meaningful even though CBIR is still suffering with several problems [4]. The process of CBIR is retrieving similar or nearby images from different or a database by matching their features or feature descriptors rather than just simply matching of N. Koteswaramma (B) JNTUA, Ananthapuramu, Andhra Pradesh, India e-mail: [email protected] Y. Murali Mohan Babu Department of ECE, Tirumala Engineering College, Narasaraopet, Andhra Pradesh, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_25

215

216

N. Koteswaramma and Y. M. M. Babu

two images. This shows the outputs, and its quality of outputs of CBIR is depending on the matching of the image feature descriptors. Generally, the features like color, texture, and shape are considered as features for the description of image descriptors [5]. Color feature is very useful and stable for direction changes and size of image complexity of background. With the help of local pattern-based descriptors, texture features are extracted. Texture feature having visible patterns contains surface structure. Shape feature is helpful in semantically characterizing the content of an image when compared to the other features. Shapes are generally defined by segmentation and edge detection methods [6]. There are several approaches for fast and reliable CBIR. There are two important points in the process of retrieval of an image using CBIR, one is construction of discriminative features, and second one is design of good similarity estimation scheme [7] and also capability of handling large scale issues. Several image descriptors have their own advantages and disadvantages [8]. In case of similarity that may not contain information of adjacent images, a new kind of CBIR is preferred which works on hypergraph and weighted adjacent structure (WAS) [9–12]. WAS improves the retrieval precision using adjacent information. Color difference histogram (CDH) and micro-structure descriptor (MSD) are combined as Co-CDHMSD method for better understanding of relationship between images [13, 14].

2 Related Work In an image, most of the information is conveyed by color, edge orientation, and texture and human eye is sensitive to both color difference and texture information contained in the digital image. As the human eye is sensitive to both color difference and texture information, it is better to combine both [15]. To achieve this, a new method is developed by combining CDH-MSD as Co-CDHMSD as represented in Eq. (1), DistCo-CDHMSD (I, j ) = t ∗ DistCDH (I, j ) + (1 − t) ∗ DistMSD (I, j )

(1)

This similarity matrix represents only information about two images, and it is not contained adjacent images of two images, which is required to describe relationship between the two images [16, 17]. This is possible now with the new method called WAS which combines the information of adjacent images for the calculation of similarity. WAS updates the original similarity matrix SM to form SM’. As shown in Fig. 1, “U” and “V ” represent nearby central images. vk is saying about the adjacent images of V. The number “n” shows the number of adjacent images [18]. W (vk,V ) can be calculated by the following formula: W (vk, V ) =

S M(vk, V ) S M(v1, V ) + S M(v2, V ) + · · · + S M(vn, V )

(2)

Color Image Retrieval with a Weighted Adjacent Structure Model

217

Fig. 1 Images U and V

Then, it is required to compute the weighted similarity SWT(U), SWT(V ) which is the similarity between all adjacent images of an image and another central image. The weighted similarity between v1,… vn and U can be calculated using the following formula shown in Eq. (3), SWT(U ) = W (v1 , V ).SM(v1 , U ) + W (v2 , V ).SM(v2 , U ) + · · · + W (vn , V ).SM(vn , U )

(3)

Then, the weighted similarity is calculated with the original similarity. The similarity matrix S′ is formed based on the final similarity between two images. S ′ (U,V ) is defined as in Eq. (4) SM′ (U, V) = (SW(U)/4) + (SW(V)/4) + (SW(U, V)/4)

(4)

After obtaining similarity matrix SM′ by WAS, the structure information of the dataset is not utilized. To achieve this, an existing soft hypergraph model is used to enhance the retrieval performance.

3 Results and Discussion Totally 2500 images have been considered for image retrieving process with five different set of images, and each has 500 images. Animal images, flower images, fruit images, vehicle images, and nature images are considered. The database images have been shown below. Figure 2 represents the 500-image database of animal pictures or images. Figure 3 represents the 500-image database of flower pictures or images. Figure 4 represents the 500-image database of nature pictures or images. In the similar fashion, fruits database of 500 images and vehicle image database of 500 images have considered. Again, these 500 images are divided into 5 sub-groups. It

218

N. Koteswaramma and Y. M. M. Babu

Fig. 2 Database of animal images

Fig. 3 Database of flower images

means 5 different types of fruits, animals, flowers, vehicles, or nature images. All the 2500 images are having 20,000 square pixels approximately. All 2500 are unique type and cover all possible types of images. Figure 5 shows query image of tiger. This image has 300*168 pixels. Figure 6 explains the histogram of CDH of query image. This histogram contains 108 features.

Color Image Retrieval with a Weighted Adjacent Structure Model

219

Fig. 4 Database of nature images

Fig. 5 Query image

Figure 7 explains the histogram of MSD of query image. This histogram has 128 features. Figure 8 explains the distance of CDH for query image. Figure 9 explains the distance of MSD for query image. Figure 10 explains the distance of Co-CDHMSD for query image. K-nearest neighbor is used to calculate similarity. Figure 11 explains the similarity and weighted similarity for query image. Calculate the weighted similarity between nearby images and central image. Figure 12 explores the hypergraph that shows the ranking. Figure 13 gives the best seven retrieved images of tiger image. Figure 14 gives the best seven retrieved images of dinosaur image.

220

Fig. 6 Histogram of CDH

Fig. 7 Histogram of MSD

N. Koteswaramma and Y. M. M. Babu

Color Image Retrieval with a Weighted Adjacent Structure Model

Fig. 8 Distance of CDH

Fig. 9 Distance of MSD

221

222

Fig. 10 Distance of Co-CDHMSD

Fig. 11 Similarity and weighted similarity

N. Koteswaramma and Y. M. M. Babu

Color Image Retrieval with a Weighted Adjacent Structure Model

Fig. 12 Hypergraph

Fig. 13 Query image and its seven retrieval images of animal images

223

224

N. Koteswaramma and Y. M. M. Babu

Fig. 14 Query image and its seven retrieval images of animal images

Figure 15 gives the best seven retrieved images of horse image. The confusion matrix is major evaluation matrix that can be done for any CBIR. From that user accuracy, producer accuracy, kappa, precision, recall, F-score, error-rate, and average rank of retrieval are calculated. Nowadays, so many machine learning, artificial learning, and deep learning techniques are available in the market to retrieve the images more accurately and faster. Still, there is much need in the speed and accurate results. It is because of the available databases in the society. This makes the people to work CBIR on different datasets, different sizes, different situations, different times, etc.

4 Conclusion This paper explains content-based image retrieval process that was already designed on available database images. The retrieval model is producing less accurate results. This process is taking less time to produce the results. In majority cases, the similar images are missing in retrieving process. The process can be developed for further good results. It gives some energy to next generation team to enhance the explained technology for much better results.

Color Image Retrieval with a Weighted Adjacent Structure Model

225

Fig. 15 Query image and its seven retrieval images of animal images

References 1. Yu S, Niu D, Zhang L, Liu M, Zhao X (2018) Colour image retrieval based on the hypergraph combined with a weighted adjacent structure. IET J 12(5):563–569 2. Radhika K, Varadarajan S, Murali Y, Babu M (2018) Multi spectral classification using cluster ensemble technique. Int J Intell Syst Technol Appl 17(1/2):55–69 3. Qi Y, Zhang G (2016) Strategy of active learning support vector machine forimage retrieval. IET Comput Vis 10(1):87–94 4. Liu GH, Yang JY, Li ZY (2015) Content-based image retrieval using computational visual attention model. Pattern Recognit 48(8):2554–2566 5. Babu YMM, Radhika K (2019) An analysis of a block matching method on single chrome images. Int J Innov Technol Explor Eng 8(8):218–220 6. Wang L, Yang B, Abraham A Distilling middle-age cement hydration kinetics from observed data using phased hybrid evolution. Soft Comput 7. Bindu Tushara D, Harsha Vardhini PA (2015) FPGA implementation of image transformation techniques with Zigbee transmission to PC. Int J Appl Eng Res 10(55):420–425 8. Shaik AS, Karsh RK, Islam M et al (2021) A review of hashing based image authentication techniques. Multimed Tools Appl. https://doi.org/10.1007/s11042-021-11649-7 9. Prakasam V, Sandeep P, Harsha Vardhini PA (2019) Snappy and video stream edge detection using labview. Int J Adv Sci Technol 28(19):197–203 10. Bindu Tushara D, Harsha Vardhini PA (2017) Performance of efficient image transmission using Zigbee/I2C/Beagle board through FPGA. In: Saini H, Sayal R, Rawat S (eds) Innovations in computer science and engineering. Lecture notes in networks and systems, vol 8. Springer, Singapore

226

N. Koteswaramma and Y. M. M. Babu

11. Gunjan VK, Shaik F, Kashyap A (2021) A tuberculosis management through ADR study, feature extraction and medical bio informatics. In: ICCCE 2020, pp 1597–1602. Springer, Singapore 12. Prasad PS, Pathak R, Gunjan VK, Rao HR (2020) Deep learning based representation for face recognition. In: ICCCE 2019, pp 419–424. Springer, Singapore 13. Kumar S, Ansari MD, Gunjan VK, Solanki VK (2020) On classification of BMD images using machine learning (ANN) algorithm. In: ICDSMLA 2019, pp 1590–1599. Springer, Singapore 14. Muralimohanbabu Y, Radhika K, Naga Jyothi D (2018) Evolutionary algorithm based extreme learning machine for retrieval of images. Int J Eng Technol UAE(IJET) 7(3.24):24, 472–475 15. Mohan Babu YM, Radhika K, ChinnaNarasimhulu G (2019) An analysis of a block matching method on different bands of monochrome images. Int J Adv Sci Technol 28(19):204–207 16. Tushara DB, Vardhini PAH (2016) Effective implementation of edge detection algorithm on FPGA and beagle board. In: 2016 International conference on electrical electronics and optimization techniques (ICEEOT), pp 2807–2811 17. Harsha Vardhini PA, Madhavi Latha M (2015) Power analysis of high performance FPGA low voltage differential I/Os for SD ADC architecture. Int J Appl Eng Res 10(55):3287–3292 18. Murali Mohan Babu Y, Subramanyam MV, Giriprasad MN (2014) A survey on de-speckling of SAR images. Int J Eng Commun Technol (IJECT) 5(4):142–144

Spam and Ham Classification by Multinomial Naïve Bayes Classification in Text Data J. K. R. Sastry, P. Harika, Trisha Dubey, and Y. Vijay Ditya

1 Introduction Spamming is the practice of delivering unsolicited advertisements to vast groups of people for the purpose of commercial, non-commercial proselytizing, or some other forbidden purpose [1–3]. Although email spam is well-known, such as instant message spam, web online classified advertising, cell phone messaging spam, spam on the Internet forum, social spam, Smartphone spam sports, TV ads, and spam file sharing [4–6]. Spamming has been focused of legislation. An individual who produces spam is called a spammer. Most email spam messages are commercial in nature [7]. Many are not only irritating, but also harmful, whether commercial or not, since they can include links that lead to malware-hosting phishing websites or sites—or have malware as file attachments [8]. Email addresses from chat rooms and blogs that steal address books from users are obtained by spammers [9]. Often, these obtained email addresses are marketed to other spammers as well.

2 Related Work A.

Multinomial Naïve Bayes Naïve says that features in the dataset are mutually independent. The Naïve Bayes will outperform the most potent choices for small sample sizes. It is used in many different fields, being relatively stable, simple to implement, fast, and precise, for example, spam filtering in messages [10].

J. K. R. Sastry (B) · P. Harika · T. Dubey · Y. Vijay Ditya Department of ECM, Koneru Lakshmaiah Education Foundation, Guntur, Andhra Pradesh, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3_26

227

228

J. K. R. Sastry et al.

Fig. 1 Multinomial Naïve Bayes classifier

Posterior Probability = B.

Conditional Probability ∗ Prior Probability Predictor Prior Probability

Stemming and Lemmatization We want to screen out spam messages using a multinomial Naive Bayes classifier. Initially, we consider eight normal messages and four spam messages (Fig. 1).

The histogram was used to measure the probability of seeing each word, provided that it was a regular message [11]. Probability (Dear|Normal) = 8/17 = 0.47. Similarly, the probability of word Friend is (Fig. 2) Probability (Friend/Normal) = 5/17 = 0.29. Probability (Lunch/Normal) = 3/17 = 0.18. Probability (Money/Normal) = 1/17 = 0.06. The possibility of the word dear we saw in the spam message given is. Probability (Dear|Spam) = 2/7 = 0.29. Similarly, the probability of word Friend isProbability (Friend/Spam) = 1/7 = 0.14. Probability (Lunch/Spam) = 0/7 = 0.00. Probability (Money/Spam) = 4/7 = 0.57.

Fig. 2 Histogram of words in spam

Spam and Ham Classification …

229

Calculate probabilities of discrete words and not probability continuous like weight or height. These probabilities are also called Likelihoods [12]. Now, let us say we have received a normal message as Dear Friend and we want to find out if it is a normal message or spam. We start with an initial guess that any message is a normal message [13]. From our initial assumptions of 8 normal messages and 4 spam messages, out of 12, 8 messages are normal messages [14]. The prior probability, in these cases, will be: Probability (Normal) = 8/(8 + 4) = 0.67. We multiply this prior with probabilities of Dear Friend that we have calculated earlier. 0.67 * 0.47 * 0.29 = 0.09. 0.09 is the probability score considering Dear Friend is a normal message. Alternatively, let us say that any message is a spam. 4 out of 12 messages are spam. The prior probability in these cases will be: Probability (Normal) = 4/(8 + 4) = 0.33. We multiply the prior probability values with probabilities of Dear Friend that we have calculated earlier. 0.33 * 0.29 * 0.14 = 0.01 0.01 is the probability score considering Dear Friend is a Spam. The probability score of Dear Friend being a normal message is greater than the probability score of Dear Friend being spam. We can conclude that Dear Friend is a normal message [15]. Naive Bayes treats all words equally regardless of how they are placed because it is difficult to keep track of every single reasonable phrase in a language.

3 Implementation of the Work A.

Holding Null Values

B.

Any real-world dataset includes a few null values. We must intervene for no paradigm can handle these NULL or NaN values on its own. First and foremost, we must determine if our dataset contains any null values. Standardization This is another step of integral preprocessing (Fig. 3). Formula for standardization z=

C.

xi − μ σ

Tokenizing Splitting down words into single elements.

230

J. K. R. Sastry et al.

Fig. 3 DataFrame

For Example, every

country

has

its

own

uniqueness

D.

Training Dataset

E.

The sample of knowledge is used to fit the model. The real dataset we use to train the model. From this data, the model sees and learns. Validation Dataset

F.

For evaluating a given model, the validation collection is used, but this is for regular assessment. We use the outcomes of the validation collection and update higher level hyperparameters. Thus, a model is influenced by the validation package, but only indirectly. It is known as the Dev packager or it can be known as production set, the validation set. It is in “development” stage of the model, and this dataset assists. Test Dataset Gold standard referred to testing the model is given by the test dataset. It is workers after a model is fully educated (using the train and validation sets). What generally, the test collection is well-selected. It includes carefully sampled data that, when used in the real world, spans the different groups encountered by the model (Fig. 4).

Fig. 4 A visualization of split

G.

Apply Multinomial Naive Bayes Algorithm For the classification of discrete characteristics, naive Bayes is appropriate .

Spam and Ham Classification …

231

For class j, word I at the frequency of the word f : Pr( j ) ∝ π j

|V | 

Pr(i| j ) Pr( j ) ∝ π j fi

i=1

|V | 

Pr(i| j ) fi

i=1

The sum of logs in order to prevent underflow: Pr( j ) = log π j +

|V | 

f i log(ti Pr(i| j ))

i=1

 Pr( j ) ∝ log π j

|V | 

 Pr(i| j )

fi

i=1

Pr( j ) = log π j +



f i log(Pr(i| j ))

j=1

One problem is that the likelihood of it happening again increases if a term appears again. We take the log of the frequency to smooth this: Pr( j ) = log π j +

|V | 

log(1 + f i ) log(Pr(i| j ))

i=1

Apply (IDF) weight on each word to take account stop terms.: ⎛

N

⎞

docn ⎟ ⎜ ⎟ ⎜ n=1 ti = log⎜ ⎟ ⎝ doci ⎠

Pr( j ) = log π j +

|V | 

f i log(ti Pr(i| j ))

i=1

Although the stop words have already been set to 0 for this particular use case, to generalize the feature, IDF implementation is being added.

4 Results This particular section provides experimental outcomes of the proposed method. First, the details of the dataset are provided. Then, predicted the performance of the

232

J. K. R. Sastry et al.

proposed model that detects if a text is spam and ham. The model is solved with the help of confusion matrix, ROC-AUC (Fig. 5). Fig. 5 Healthy message indication

A.

Uncertainty matrix: It can be used to compute precision, sensitivity (aka recall), specificity, and accuracy (Fig. 6).

Fig. 6 Uncertainty matrix

We compute make use of the specified formulae to explain the accuracy for offered model. Accuracy is the ability to determine the correctness or closeness of personality categorization. The formula for accuracy can be given by Accuracy = (TP + TN)/(TP + TN + FP + FN) Precision mention to the correctness of more than two values. Precision = TP/(TP + FP)

Spam and Ham Classification …

233

Specificity is classified as percentage of negatives that are specifically identified in a binary classification test and can be calculated by (Fig. 7) Fig. 7 Accuracy of the developed model

Specificity = TN/(TN + FP) Accuracy of the model that says the text is spam or ham is 97% (Fig. 8).

Fig. 8 ROC-AUC bend

The AUC of the model of multinomial naive Bayes is 0.98, which is closer to 1. Based on the ROC-AUC curve, it is clear that the AUC of the model of multinomial naive Bayes is 0.98, which is closer to 1.

234

J. K. R. Sastry et al.

5 Conclusion This paper shows a simple machine learning method and created an output assessment for it using the spambase dataset. The results with the hypothetical power, as well as the limitations of every technique, are also precise. The ability of individual classifiers to accurate identify text message as spam, ham is an interesting extension of this working. When constructed using the multinomial naive Bayes classifier, our model had a 97% precision, a 0.98 ROC-AUC curve, and a 0.96 GINI coefficient rating.

References 1. Developer Policy Center—intellectual property, deception, and spam. play.google.com. Retrieved 01 May 2016 2. "Spam”. Merriam-Webster Dictionary (definition & more). 31 Aug 2012. Retrieved 05 July 2013. 3. "The Definition of Spam”. The Spamhaus Project. Retrieved 03 Sept 2013. 4. Gyöngyi Z, Garcia-Molina H (2005) Web spam taxonomy (PDF). In: Proceedings of the first international workshop on adversarial information retrieval on the web (AIRWeb), 2005 in the 14th international world wide web conference (WWW 2005), May 10, (Tue)–14 (Sat) 5. Saab SA, Mitri N, Awad M (2014) Ham or Spam? A comparative study for some contentbased classification algorithms for email filtering. In: 17th IEEE Mediterranean electrotechnical conference, Beirut, Lebanon, 13–16 Apr 2014 6. Rajput AP, Athavale V, Mittal S (2019) Intelligent model for classification of SPAM and HAM. Int J Innov Technol Explor Eng (IJITEE) 8(6S) 7. Metsis V, Androutsopoulos I, Paliouras G (2006) Spam filtering with Naive Bayes—which Naive Bayes? In: CEAS 2006 third conference on email and antispam, July 2006 8. Gupta A, Mohan KM, Shidnal S (2018) Spam filter using Naïve Bayesian technique. Int J Comput Eng Res (IJCER) 9. Koppula VK, Atul N, Garain U (2009) Robust text line, word and character extraction from telugu document image. In: 2009 2nd international conference on emerging trends in engineering and technology, ICETET 2009 10. Narayana VA, Premchand P, Govardhan A (2009) A novel and efficient approach for near duplicate page detection in web crawling. In: 2009 IEEE international advance computing conference, IACC 2009 11. Merugu S, Tiwari A, Sharma SK (2021) Spatial–spectral image classification with edge preserving method. J Indian Soc Remote Sens. https://doi.org/10.1007/s12524-020-01265-7. ISSN 0255-660X 12. Merugu S, Reddy MCS, Goyal E, Piplani L (2019) Text message classification using supervised machine learning algorithms. In: Kumar A, Mozar S (eds) ICCCE 2018. Lecture notes in electrical engineering, , vol 500, ISSN 1876-1100. Springer, Singapore 13. Devadasu G, Sushama M (2016) A novel multiple fault identification with fast fourier transform analysis. In: 1st International conference on emerging trends in engineering, technology and science, ICETETS 14. Thanuja Nishadi AS (2019) Text analysis: Naïve Bayes algorithm using Python JupyterLab. Int J Sci Res Publ 15. Gupta P, Dubey RK, Mishra S (2019) Detecting spam emails/sms using Naive Bayes and support vector machine. Int J Sci Technol Res 8(11)

Author Index

A Abdul Subhani Shaik, 163 Agilesh Saravanan, R., 39 Anuradha, B., 203 Aruna Sri, P. S. G., 9, 195 Arvind Yadav, 113

Gowri Priya, 39

H Harika, P., 227 Haripriya Mishra, 113 Harshitha, A., 87 Himabindu, K., 157

B Bhargavi, P., 101 J Jyosthna, R., 65 C Chakradhar, B., 123 Chakravarthy, V. V. N., 47 Challa MadhaviLatha, 185 Challa Venkata Pranith, 29 Ch. Amarnath, 71 Chavala Lalithya Rao, 9 Ch. Mallikarjuna Rao, 185 Ch. Manikanta Uma Srinivas, 87

D Deepak Kumar, B. P., 171 Devendra Joshi, 113

E Eswar Sai, M., 87

G Gopi Krishna, P., 87, 101 GopiRam, V., 47 Govardhan Reddy, M., 79

K Kameswara Rao, M., 171 Kiran Bhavya, N., 55 Kiran, K. V. D., 29 KishoreBabu, K., 47 Kolli Sai Kiran, 29 Koteswaramma, N., 215 Krishnamraju, K., 123 Krishnaveni Kommuri, 87 Kumar, K. T. P. S., 131 Kurapati Sainath Raju, 9

L Lakshman Pappula, 131 Lakshmi Anusha, M., 19 Lavanya, K., 123

M Mallesh, S., 147 Medha Swapnika Kidambi, 19

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 V. K. Gunjan et al. (eds.), Cybernetics, Cognition and Machine Learning Applications, Algorithms for Intelligent Systems, https://doi.org/10.1007/978-981-19-1484-3

235

236 Mohammed Ali Hussain, 123 Murali Mohan Babu, Y., 215

N NageswaraRao Lavuri, 163 Nallamalapu Pratapreddy, 131 Narayana, V. A., 9

P Panduranga, V., 147 Pasupathy, S., 1 Pavan Kumar, R., 65 Potharaju Vinay Kumar, 131 Pragati Mishra, 9, 113 Prasad, M. V. D., 19 Praveen Sai, 39 Pulivarthi Goutham, 29 Pushpa Rani, K., 157

R Rakesh Kumar, Y., 141 Ranga Sai Kiriti, V., 195 Revanth, A., 65 Rohit, S., 101

S Sai Bhaskar Reddy, B., 79 Sai Deepak, N., 195 Sai Kalyan, M. V., 131 Sai Nishanth, 39 Sai Prasad, K., 1 SaiSaketh, K., 55 Sanjan Miller, P., 79 Sanjay Vishnoi, 113 Sanjitha, K., 147 Santhosh Kumar, V., 163 Sastry, J. K. R., 227 Satyanarayana Goud, P., 141

Author Index Shashank, S., 157 Sheelam Pravalika, 141 Siva Ganga Prasad, M., 101 Soujanya, K. L. S., 185 Sreenivas, B., 203 Sree Vardhan Cheerla, 47, 55 Sree Varsha, K., 123 Sridevi Sakhamuri, 185 Sridhar, M., 65, 71, 79 Srivastav, K. D. S. R. S. H., 19 Subba Reddy, V., 55 Surendra Kumar Bitra, 65, 71, 79 Surya Teja, K., 71 Sushma Swaraj, N., 157 Swathi, M., 185 Syed Inthiyaz, 55

T Tanmayi, B., 101 Teja Kiran Kumar, M., 19 Trisha Dubey, 227

V Valiveti Lohya Sujith, 29 Vamseekrishna, A., 101 Vankayalapati Sahiti, 131 Vara Prasad, B., 157 Vasanth Kumar, 39 Veera Bhadra, K., 195 Venkateswara Reddy, R., 171 Venkateswara Rao, K., 171 Venkateshwarla Rama Raju, 203 Venu Adepu, 195 Vijay Ditya, Y., 227 Vinay Kumar, M., 55

Y Yoga Sasidhar Reddy, S., 71