Multicriteria Decision Aid and Resource Management: Recent Research, Methods and Applications (Multiple Criteria Decision Making) 3031348915, 9783031348914

Table of contents :
Preface
Contents
Chapter 1: The Multicriteria Features of Environmental Footprint Assessment Methodology for Refractory Materials in the Circul...
1.1 Introduction
1.2 Environmental Footprint: An Instrument for Sustainability
1.3 Multicriteria Features of Environmental Footprint Assessment
1.4 Conclusions
References
Chapter 2: Fire Disaster Recovery and Resource Allocation Enabled by Firefighters´ Sustainable UAV Technology in Smart Cities
2.1 Introduction
2.2 Related Work
2.3 Control System Architecture
2.4 Evaluation Parameters
2.5 Evaluation Metrics
2.6 Use Cases
2.7 Experiments
2.8 Results and Discussion
2.9 Conclusions and Future Work
References
Chapter 3: Resource Management: A Bi-Objective Methodological Approach for Routing in Crisis Situations
3.1 Introduction
3.2 Literature Review
3.3 Methodology
3.4 Illustration
3.5 Conclusions
References
Chapter 4: Urban Space Quality Evaluation Using Multi-Criteria Decision Analysis-Based Framework
4.1 Introduction
4.2 Literature Review
4.2.1 Public Space and Contemporary Challenges
4.2.2 Quality Evaluation
4.2.3 Urban Space and Multi-criteria Analysis
4.3 Methodology
4.3.1 The Underlying Principles of Multi-Criteria Analysis
4.3.2 Setting up the Decision Problem from an Urban Planning Perspective
4.3.3 Setting up the Decision Problem from MCDA Perspective: Toward the Construction of the Decision Performance Table
4.3.4 The Proposed MCDA Methodology
4.3.5 Numerical Example
4.4 Conclusions
References
Chapter 5: Multicriteria Disaggregation- Aggregation Approach for the Evaluation of Warm Water Lakes
5.1 Introduction
5.2 Environmental Issues of Lakes´ Water
5.2.1 Point Sources of Water Pollution
5.2.2 Other Sources of Water Pollution
5.3 Categorization of Lakes Based on Their Nutrient Content
5.3.1 pH
5.3.2 Phosphorus
5.3.3 Nitrogen
5.3.4 Depth of Secchi Disc
5.3.5 Chlorophyll-a in Phytoplankton
5.4 Methodology
5.4.1 The UTA II Method of the Multicriteria Disaggregation: Aggregation Approach
5.4.2 Methodological Approach for Ecological Classification
5.5 Case Study of Warm Lakes´ Water Classification
5.6 Conclusion and Perspectives
References
Chapter 6: Agricultural Water Management in the Context of Water-Energy-Land-Food NEXUS
6.1 Introduction
6.1.1 The Water-Energy-Land-Food Nexus (WELF Nexus) Framework
6.1.2 Agricultural Water Management Challenges in the Context of WELF Nexus
6.1.3 Applications of Multi-criteria Decision Aid (MCDA) in Water and Agricultural Management
6.1.4 Demonstration Case
6.2 Methodology
6.2.1 Developing the Decision Performance Table
6.2.2 Selecting an Evaluation Model
6.2.3 Performing Preference Elicitation and Modelling
6.2.4 Developing the Marginal Value Functions Using the MIIDAS System
6.2.5 Assessing the Points of View and Criteria Weights Using the WAP Technique
6.2.6 Evaluation of Alternative Agricultural Water Management Schemes
6.3 Demonstration Case Results
6.3.1 Developing the Decision Performance Table
6.3.2 Alternatives
6.3.3 Points of View and Criteria
6.3.4 Evaluation of Performance
6.3.5 Performing Preference Elicitation and Modelling
6.3.6 Developing the Marginal Value Functions Using the MIIDAS System
6.3.7 Assessing the Points of View and Criteria Weights Using the WAP Technique
6.3.8 Evaluation of Alternative Agricultural Water Management Schemes
6.4 Discussion
6.4.1 Lessons Learnt from the Application of the Proposed MCDA Framework
6.4.2 Key Trade-Offs in the Demonstration Case
6.5 Conclusions
References
Chapter 7: Recommending Open Educational Resources Using Weighted Linear Combination
7.1 Introduction
7.2 Description of Characteristics
7.2.1 Description of the WLC-Based Technique
7.3 Evaluation Results and Discussion
7.4 Conclusions
References
Chapter 8: User Comments as a Resource to Rank with Multiple Criteria: The Case of TripAdvisor Athens´s Restaurants
8.1 Introduction
8.2 Literature Review
8.3 Online Restaurant Reviews
8.4 Topic Modeling
8.5 Multi-criteria Decision-Making
8.6 Methodology
8.6.1 Research Method
8.6.2 Proposed Ranking Mechanism
8.7 Experimental Results
8.7.1 Data Availability and Setting
8.7.2 LDA Setting
8.7.3 Ranking All Restaurants
8.8 Discussion
8.9 Conclusions
References
Correction to: Multicriteria Disaggregation- Aggregation Approach for the Evaluation of Warm Water Lakes
Correction to: Chapter 5 in: A. Spyridakos (ed.), Multicriteria Decision Aid and Resource Management, Multiple Criteria Decisi...

Citation preview

Multiple Criteria Decision Making

Athanasios Spyridakos   Editor

Multicriteria Decision Aid and Resource Management Recent Research, Methods and Applications

Multiple Criteria Decision Making Series Editor Constantin Zopounidis, School of Production Engineering and Management, Technical University of Crete, Chania, Greece

This book series focuses on the publication of monographs and edited volumes of wide interest for researchers and practitioners interested in the theory of multicriteria analysis and its applications in management and engineering. The book series publishes novel works related to the foundations and the methodological aspects of multicriteria analysis, its applications in different areas in management and engineering, as well as its connections with other quantitative and analytic disciplines. In recent years, multicriteria analysis has been widely used for decision making purposes by institutions and enterprises. Research is also very active in the field, with numerous publications in a wide range of publication outlets and different domains such as operations management, environmental and energy planning, finance and economics, marketing, engineering, and healthcare. This series has been accepted by Scopus.

Athanasios Spyridakos Editor

Multicriteria Decision Aid and Resource Management Recent Research, Methods and Applications

Editor Athanasios Spyridakos Business Administration Department University of West Attica, Faculty of Business, Economical and Social Sciences Aigaleo, Athens, Greece

ISSN 2366-0023 ISSN 2366-0031 (electronic) Multiple Criteria Decision Making ISBN 978-3-031-34891-4 ISBN 978-3-031-34892-1 (eBook) https://doi.org/10.1007/978-3-031-34892-1 © Springer Nature Switzerland AG 2023, corrected publication 2023 This work is subject to copyright. All rights are reserved 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 Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

Resources management constitutes a major issue of modern times whether it concerns human or material resources. The recent economical status and the dramatically disrupted environmental conditions simultaneously along with the population explosion emerge the rational and better utilization of resources at local, national and global level. The problem has been grown on great extent while the inexhaustible use of earth resources has created undesirable situations in the environment and erected questions about the evolution of life on earth. Multicriteria decision aid analysis includes methodological tools which aim to support the decision-making at operational and executive level to all aspects of the economic and social life, furthermore to these crucial issues related to the resources planning and management. Multicriteria Decision Aid Analysis provides the methodological tools to handle unstructured decision-making problems where there are many conflicting criteria to be considered, the parameters are liquid and there is a high degree of uncertainty of the decision impacts. This book includes selected research works related to the exploitation of the Multicriteria Decision Aid Analysis on sensitive resource management case studies and problems. The recycling or re-use of the waste is considered one of the most important activities to the resource management while it is strictly related with the sustainability, supports the rational use of the resources, reflects positive to the environmental impacts reduction for a better life. The first research work examines the utilization of the MCDA framework to the Environmental FootPrint Assessment focused on the refractory waste. The aim of the proposed approach is to support the decisionmaking processes for actions leading to the reduction of the environmental impacts concerning the crucial resources such as mine and water. The second research work is related to the management of the Unmanned Aerial Vehicles (UAV) in order to support the immediate response to fire disasters and preventing the escalation of the fire. The UAVS are also used for the better utilization of the available resources (human, vehicles, airplanes, water, etc.) which can lead to the minimization of the disaster impacts and save lives and infrastructures. v

vi

Preface

The fire disaster case is also examined in the third chapter while the effective management of the health related resources (ambulances, medicine staff and hospital units) is examined and a companion of bi-criterion and dynamic programming approach is proposed for the immediate response and handling patients taking into consideration the severance of their injuries. The urban quality spaces are examined and analysed in the next chapter. Actually the Weights Assessment through Prioritization (WAP) approach in companion with the TOPSIS method is employed in order to evaluate the urban spaces quality. The aim of this research is to support decision-makers to identify the strengths and weaknesses of the community and finally to lead to the better and sustainable management of these resources. The next two research works are focused on the water management. The first one proposed a multicriteria methodological frame for the environmental evaluation of the lakes’ water. The disaggregation–Aggregation UTA II method is utilized in order to construct an additive value system providing more inclusive and detailed way to discriminate the lakes’ water according to their environmental situation. The sixth chapter involved the MCDA framework for the management of the water for agricultural purposes. The Weights Assessment through Prioritization (WAP) Method within Multi Attribute Utility Theory techniques is used into the framework of the NEXUS methodological approach in order to support decision-making situations concerning the water use for agricultural activities. The seventh research work examined the rational and effective utilization of open educational resources used in learning activities while the precise and the completeness constitute two crucial factors. A weighted linear combination model is used into a recommender system taking into consideration the preference and features of the learners as well as the learning materials characteristics. Finally, the eight chapter proposed a mechanism for ranking assessment, which relies on the qualitative characteristics by incorporating Latent Dirichlet Allocation and multi-criteria decision-making methodological frameworks. The ranking mechanism is used for the TripAdvisor’s user comments on resources for a region. Athens, Greece

Athanasios Spyridakos

Contents

1

2

3

4

5

6

7

The Multicriteria Features of Environmental Footprint Assessment Methodology for Refractory Materials in the Circular Economy: Issues, Perspectives, and New Directions . . . . . . . . . . . . . . . . . . . . . Athanasios Spyridakos, Dimitrios E. Alexakis, Isaak Vryzidis, Nikos Tsotsolas, George Varelidis, and Efthimios Kagiaras

1

Fire Disaster Recovery and Resource Allocation Enabled by Firefighters’ Sustainable UAV Technology in Smart Cities . . . . . . . . Theodoros Anagnostopoulos and Yannis Psaromiligkos

19

Resource Management: A Bi-Objective Methodological Approach for Routing in Crisis Situations . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stamatios Vasalakis and Athanasios Spyridakos

33

Urban Space Quality Evaluation Using Multi-Criteria Decision Analysis-Based Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Athina Mela, Isaak Vryzidis, George Varelidis, and Nikolaos Tsotsolas

59

Multicriteria Disaggregation- Aggregation Approach for the Evaluation of Warm Water Lakes . . . . . . . . . . . . . . . . . . . . . . . . . . Dimitrios E. Alexakis, Isaak Vryzidis, and Athanasios Spyridakos

85

Agricultural Water Management in the Context of Water–Energy– Land–Food NEXUS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Alexander Psomas, Isaak Vryzidis, Nikolaos Tsotsolas, and Maria Mimikou Recommending Open Educational Resources Using Weighted Linear Combination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Christos Troussas, Akrivi Krouska, Panagiotis Douros, and Cleo Sgouropoulou

103

137

vii

viii

8

Contents

User Comments as a Resource to Rank with Multiple Criteria: The Case of TripAdvisor Athens’s Restaurants . . . . . . . . . . . . . . . . Dimitris Novas, Dimitris Papakyriakopoulos, Elissavet Kartaloglou, and Anastasia Griva

Correction to: Multicriteria Disaggregation- Aggregation Approach for the Evaluation of Warm Water Lakes . . . . . . . . . . . . . . . . . . . . . . . Dimitrios E. Alexakis, Isaak Vryzidis, and Athanasios Spyridakos

145

C1

Chapter 1

The Multicriteria Features of Environmental Footprint Assessment Methodology for Refractory Materials in the Circular Economy: Issues, Perspectives, and New Directions Athanasios Spyridakos, Dimitrios E. Alexakis, Isaak Vryzidis, Nikos Tsotsolas, George Varelidis, and Efthimios Kagiaras

Abstract Refractory materials are extendedly used in the infrastructure of heavy industries to face high-temperature conditions. Different refract products are used across the various sectors depending on the production condition and requirements. A considerable quantity of 25–30 million tons of refractory materials waste is produced every year worldwide with an increasing trend. The treatment of the spent refractory materials includes landfill, clearness, and recycling activities or a combination of them. This research examines the environmental footprint methodologies used worldwide and the impacts of the production and use of refractory materials in sustainable development. This is considered vital for heavy industries since it is linked with the footprint assessment of their products. Also, a methodological framework is proposed for assessing the Environmental Footprint based on the principles and requirements of the Disaggregation–Aggregation Multicriteria

A. Spyridakos (✉) · N. Tsotsolas Laboratory of Business Informatics and Quantitative Methods, Department of Business Administration, School of Administrative, Economics and Social Sciences, University of West Attica, Athens, Greece e-mail: [email protected]; [email protected] D. E. Alexakis · I. Vryzidis Laboratory of Geoenvironmental Science and Environmental Quality Assurance, Department of Civil Engineering, University of West Attica, Athens, Greece e-mail: [email protected]; [email protected] G. Varelidis Laboratory of Urban Planning and Architecture, Department of Civil Engineering, University of West Attica, Athens, Greece e-mail: [email protected] E. Kagiaras R&D Department, Mathios Refractories S.A., Athens, Greece e-mail: [email protected] © Springer Nature Switzerland AG 2023 A. Spyridakos (ed.), Multicriteria Decision Aid and Resource Management, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-34892-1_1

1

2

A. Spyridakos et al.

Decision Aid approach to provide a more structural and accurate way to evaluate the environmental impacts.

1.1

Introduction

Through applying alternative use cycles, the circular economy (CE) aims at reducing dependency on (new) natural resource extraction while expanding the time that resources are spent within the technology sphere (Tan & Lamers, 2021). In essence, this approach of CE imitates Earth’s naturally circular systems. CE envisions a balance between industry and the environment, ensuring that human civilisation will continue to thrive in the long run. All waste is designed to become reusable material through different processes. Many researchers have studied the aspects raised by the two main pillars of CE: the technical and biological models. Ferreira et al. (2015), Horckmans et al. (2019), Fang et al. (1999), Hanagiri et al. (2008), Poirier et al. (2017), and Arianpour et al. (2007) dealt with refractory characterisation and recycling. Moreover, Vlachokostas et al. (2021) suggested a framework supporting the decision-making for biodegradable waste management. On the other hand, Kyriakopoulos et al. (2019) concluded that there are still many issues leading to the negative environmental impacts arising from CE’s fundamental assumptions. Synthetic ceramics and refractories are essential to modern society. They own a large share of the global market for ceramic materials. The production volumes of ceramic refractories are significantly high, as they are essential mainly for the steel industry (Horckmans et al., 2019). Therefore, information about refractory applications and sources of refractory raw materials is necessary for a sustainable operation. Technical ceramics offer unique properties for many different applications. These materials can be used in applications where other materials are not suitable. Technical ceramics also enable the emergence of new technologies in many industries (Deneen & Gross, 2010). Building materials and refractories represent the largest share of the European ceramics market. This is due to the increasing need for refractory materials for hightemperature processes and the continued demand for ceramic tiles for building construction. The high value and production volume are because these materials have been used as an affordable alternative to metal building materials. The consumption of refractory ceramics is very high, especially in industries with hightemperature processes and high production volumes. The steel industry is the largest. The paper is organised as follows: Section 1.1 includes the introduction, which provides an overview of the current literature related to the specific topic. Section 1.2 provides a review of the idea of the environmental footprint as an instrument for sustainability. Section 1.3 presents the fundamentals of the ReCiPe method for the environmental footprint assessment and analyses its multicriteria features (European Commission, 2013). In this section, issues on the multicriteria assessment of environmental footprint are analysed, and a methodology is proposed that addresses

1

The Multicriteria Features of Environmental Footprint. . .

3

them. Section 1.4 concludes the study’s main findings and offers new avenues for future research.

1.2

Environmental Footprint: An Instrument for Sustainability

William Rees and Mathis Wackernagel first brought the idea of leaving an ecological footprint to the scientific community’s attention (Bortsie-Aryee & Gabriel, 2020). According to many researchers (Rees, 1999; Van den Bergh & Verbruggen, 1999; Wackernagel & Rees, 1998; Wiedmann & Barrett, 2010), a methodology known as the ecological footprint can be utilised to evaluate the influence that human activities have on the natural environment. It is a popular indicator of the number of natural resources required to support human systems. In other words, it is an indicator that compares the level of consumption of natural resources with the amount of productive land and sea area available to support this consumption. This consumption can be supported by productive sea and land regions (Wiedmann & Barrett, 2010). Ecological footprint demonstrates the interdependence of human populations (considered from the perspectives of individuals, nations, and the entire world) on their natural environments (Monfreda et al., 2004; Wackernagel & Rees, 1997, 1998). Researchers who apply the ecological footprinting methodologies note that the ecological footprint may be separated from other sustainability indicators in two significant ways (Hoekstra, 2009). These ways are: (a) it can be measured and (b) it can be compared to other footprints. It achieves this in two ways: (a) it quantifies the influence of mankind using a single unit and (b) it quantifies the impact of humanity concerning the carrying capacity of the Earth or the capability of the Earth to sustain life. Concerns about the decreasing carrying capacity of the Earth have led to the rise in popularity of the concept of ecological footprints. These footprints have proven to be an effective baseline for acknowledging and quantifying the current impact that anthropogenic activity has on the natural environment, which is necessary before developing solutions that may improve this impact. Since then, ecological footprints have also been utilised as a benchmark, quantifying and communicating recommended consumption levels for society. One of the primary obstacles to achieving environmental sustainability is overcoming the effects of carbon emissions caused by human activities and how those emissions contribute to climate change. Numerous studies have been performed to quantify the environmental performance of a product, industry, town, or region by using environmental Life Circe Assessment (LCA) in conjunction with the carbon footprint, which serves as an environmental sustainability indicator. Various LCA methodologies have been proposed to evaluate the environmental effects of materials throughout their life cycle, from the extraction of resources through production, consumption, and disposal of waste. These methods include the bottom-up approach-based LCA methodology and the top-down input–output-based LCA

4

A. Spyridakos et al.

methodology. Both of these methodologies present benefits and drawbacks, particularly concerning the system’s boundaries and the manufacturing processes’ specifics. However, we need both methods to delineate the carbon implications of anthropogenic activities at different scales, such as regional development, company operations, and individual consumption. This is because the carbon implications of human activities vary greatly depending on the scale at which they are studied. In addition, a win–win analysis between the carbon footprint and the water footprint or the land footprint may provide managers and stakeholders with crucial information that can help them achieve overall environmental sustainability. In order to make our communities more resilient and sustainable, it is necessary to gain a better understanding of, and work towards reducing, the environmental impact caused by consumers, who are both key decision-makers and stakeholders. Tools that measure a person’s “environmental footprint” connect the day-to-day resource consumption of consumers with the impact their choices have on the environment. The purpose of nitrogen footprints is to address the challenge of utilising nitrogen effectively for the production of food and energy, while at the same time minimising the adverse effects on human and environmental health. The widespread application of nitrogen footprint tools, the goals derived from them, and the actual footprint reductions have contributed to their success. One of the most important policy priorities on a global scale is currently the enhancement of environmental quality while simultaneously increasing resource productivity. The intensive consumption of material resources harms environmental quality, manifesting in the form of biodiversity loss, climate shifts, intensified water contamination, and the depletion of natural resources. In addition to this, the rising utilisation of natural resources raises the question of the availability of these resources for nations (Sinha, Sengupta, & Saha, 2020; Sinha, Shah, et al., 2020). In light of this, one of the most hotly contested issues in ecology in recent years has been consuming the ecosystem’s material resources. According to the Energy Transitions Commission (2017), the consumption of heavy material resources is responsible for a staggering 30% of world emissions. The improvement of resource efficiency and the guarantee of sustainable management (reduce, reuse, and recycle) of materials is the primary focus of growth policy in every region of the world (Sinha et al., 2021; Sinha, Shah, et al., 2020). Not only is environmental degradation caused by the extraction and processing of materials consumed, but it is also caused by existing material stocks, which are responsible for a significant portion of the world’s emissions (Krausmann et al., 2020). As a result, the mechanisms driving such shifts should be of great concern to decision-makers. According to Bringezu and Bleischwitz (2009), the overall use of material resources per capita annually varies between 40 and 50 tonnes in most industrialised economies. Economic growth and human well-being must be decoupled from resource consumption to achieve sustainable development, which would otherwise be impossible (Khan et al., 2020; Krausmann et al., 2020). Decoupling improves not only environmental quality by reducing the number of resources required by economies and the environmental effects associated with those resource demands but also

1

The Multicriteria Features of Environmental Footprint. . .

5

maintains economic development by increasing material supplies (Sharif et al., 2020). Understanding how resources move through economies and how material resource efficiency can be increased by utilising specific technical changes and adapting to technological advancement is critical for successfully implementing policy implications (Khan et al., 2020). This understanding is necessary for making successful policy recommendations. As a result, appropriate consumer policy on both the national and international levels is required for sustainable resource management with different strategies depending on the level of resource consumption in an economy. The extent and scope of the adverse effects on the environment that are caused by different consumption patterns have been investigated in previous research. Mobility (vehicles and aeroplanes), food and energy consumption have been found to be the primary causes of the environmental impressions accounting for 80% of the ecological footprints (Huppes et al., 2006; Tukker & Jansen, 2006). These environmental perceptions are the outcome of global warming, leading to storms, droughts, and biodiversity loss (Abubakar & Bununu, 2020).

1.3

Multicriteria Features of Environmental Footprint Assessment

The Product Environmental Footprint (PEF), according to the Commission Recommendation 2013/179/EU, is a multicriteria measure of the environmental performance of a good or service throughout its life cycle (European Commission, 2013). The primary purpose of the Environmental Footprint (EF) assessment is to reduce the impacts of goods and services considering supply chain activities (extraction of raw materials, transport, production and use, and final waste management) (Čuček et al., 2012; Matuštík & Kočí, 2021). Related studies are suggesting a broader set of relevant environmental performance criteria using a life-cycle approach by taking into account the spectrum of resource flows and environmental interventions associated with a product or organisation (Fang et al., 2013, 2014, 2016; Wiedmann & Barrett, 2010). The EF Assessment process can be divided into five stages (Fang & Heijungs, 2015) following the waterfall approach, where starting one stage requires completing the previous one (Fig. 1.1). The first stage (Goals Definition) concludes with a study concerning the (goal setting) identification of the goals for the EF assessment and the reasons for carrying out the analysis, as well as the involved partners and their roles. The second stage (Scope Identification) includes activities to define the analytical specifications of the EF assessment and the clarification of the questions related to the implementation of the study, such as the functional unit on which will be based the EF analysis and the parameters of the analysis (i.e. What we have to measure? Why do we have to measure it? How many of the production units will be used to assess the environmental impacts? What is the expected quality level? What is the lifetime of the product?)

6

A. Spyridakos et al.

Define Goals

Define Scope

Create the Resource Use and Emissions Profile

Environmental Footprint Review

Conduct Environmental Footprint Impact Assessment Environmental Footprint Interpretaon and Reporng

Fig. 1.1 Environmental Footprint Assessment—Waterfall process (European Commission, 2013)

Following the previous preliminary stages, the profiles of all resource inputs (use) and emissions resulting from a functional unit’s production are compiled, i.e. the resource use and emissions are identified and estimated for the whole production cycle, from resource mining to disposal or recycling at the end of the product lifetime. Data collected from the production process, other related studies and publications, official statistics and databases are used for the detailed recording of the resource inputs and emissions. The accuracy, precision, and uncertainty of the recorded data are examined while the process is reviewed in this stage to ensure the reliability of the study. The next and fourth stage leads to the estimation of the EF through the following steps. The first step in this stage is categorisation, which refers to assigning the inputs and emissions to the 16 categories of environmental impacts (Table 1.1). This is achieved by following specific rules that are obtained for every product or process. The characterisation step follows, where the amounts that the categorised inputs and emissions contribute to the effects of the respective category are estimated. For quantifying these effects, every resource input or emission is converted to an equivalent measure, which has been formally identified for every impact category. Also, the aggregated values are multiplied by the respective multipliers (factors) corresponding to each impact category. These multipliers correspond to surrogates or resources and represent the intensity of the effects of a surrogate

1

The Multicriteria Features of Environmental Footprint. . .

7

Table 1.1 Impact categories, units, and normalisation factors

No 1 2 3 4 5 6

7

8 9 10 11 12 13

14

Impact category Climate change Ozon depletion Ecotoxicity for aquatic freshwater Human toxicity— cancer effects Human toxicity, non-cancer effects Particulate matter/ respiratory inorganics Ionising radiation— human health effects Photochemical ozone formation Acidification

Unit Kg CO2 – eq. Kg CFC-11 – eq. CTUe

Eutrophication— terrestrial Eutrophication— aquatic Resources depletion—water Resource depletion—mineral and metals Land transformation

Mol N – eq.

CTUh CTUh Kg PM 2.5 – eq.

Domestic 4.6E+12 1.08E+7 4.36E +12 1.84E +04 2.66E +05 1.90E +09

Normalisation factor per person 9.22E+3 2.16E-02 8.74E+03

Overall Robustness Very high Medium Low

3.69E-05

Low

5.33E-04

Low

3.80E+00

Very high

Kg U235 – eq. (to air)

5.64E +11

1.13E+03

Medium

Kg NMVOC – eq.

1.58E +10 2.36E +10 8.76E +10 7.41E +08 4.06E +10 5.03E +07

3.17E+01

Medium

4.73E+01

High

1.76E+02

Medium

1.48E+00

1.01E-01

Medium to low Medium to low Medium

7.48E+04

Medium

Mol H+ – eq.

Kg P – eq. (freshwater) Kg N – eq. (marine) m3 water use related to local scarcity of water Kg antinomy (Sb) – eq. Kg (deficit)

3.74E +13

8.14E+01

relative to a reference surrogate (equivalent). There are two levels regarding characterisation: (a) the midpoint and (b) the endpoint (Fig. 1.2). At the midpoint level, the calculation of the factors is focused on each environmental impact, while at the endpoint level, the effect on the three complex and dominant areas of protection (human health, ecosystem, resource scarcity) is estimated by the aggregation of the MidPoint level results. The most common method for environmental impact assessment is the ReCiPe method (Huijbregts et al., 2017) which was created by RIVM, Radboud University Nijmegen, CML, and PRé Consultants. It is also important to highlight that many variations of the ReCiPe Method have been developed. The two processes of normailization and weighting, non manadatory but useful for the EF estimation, follow the steps of classification and characterisation. The results of the Life Cycle Impact analysis are multiplied by normalisation factors to enable the results to be comparable in terms of their contribution to the

Categorisation

Eutrophication - terrestrial

…

Characterisation

Oil / coal energy cost

Increased extraction costs

Damage to marine species

Damage to terrestrial species

Damage to freshwater species

Increase in malnutrition

Increase in other diseases / causes

Increase in various types of cancer

Increase respiratory disease

Damage pathways

Fig. 1.2 Pathways for midpoint and endpoint levels covered in the ReCiPe method (Huijbregts et al., 2017)

Land Transformation

Resource Depletion – mineral, fossil

Resource Depletion - Water

Eutrophication - aquatic

Photochemical Ozone Formation

Acidification

Ionizing radiation

Particulate Matter /Respiratory Inorganics

Human toxicity (non-cancer)

Human toxicity (cancer)

…

O3

Pb

As

CH4

CO2

Emissions & Inputs

Freshwater ecotoxicity

Ozone depletion

Climate Change

Midpoint impact category

Damage to resource availability

Damage to ecosystems

Damage to human health

Endpoint area of protection

8 A. Spyridakos et al.

1

The Multicriteria Features of Environmental Footprint. . .

9

Environmental Footprint based on the reference units. The Normalised Factors (Table 1.1) used in every case are depended on the region, the scope of the analysis, and the occasion under examination. The normalised results of the impact analysis are weighted to conclude with an index that captures the total environmental impact. The weighting allows setting priorities in the individual categories of the environmental impacts depending on the purpose and aim of the EF assessment application. The Joint Research Center of the EU has proposed (non-binding) specific weighting factors resulting from the application of multicriteria approaches in the context of a multi-participant technical study in the determination of gravities (Joint Research Centre (European Commission) et al., 2018). Regarding the Refractory material, the abovementioned process for EF assessment will be utilised, and it will be based on the Life Circle (LC) for the production and use of the refractory materials described below (Fig. 1.3). Specifically, it includes: (a) Raw material extraction: Mining activities in the case of Greek Industries are carried out in foreign countries, mainly China. It is worth mentioning that part of the raw materials comes from recycling used refractory products and porcelain. Inputs in this process are the mining products and the water used, while emissions are produced mainly by energy consumption (fuels, electricity). Related impact categories are Climate Change, Resource Depletion and Resource Depletion—Water. (b) Raw material transformation: The raw material is transferred to the industries in various ways. Regarding Greek sectors, this is mainly achieved by maritime transportation. The raw material extracted in China must be transported far more than 6000 miles burdening the environment with Carbon dioxide from the ship’s and vehicles’ fuel. This is related to the Climate Change impact category. (c) Preparation phase: It includes crushing, milling, and sieving activities, which require special industrial machinery equipment. The emissions are produced by water consumption, which is needed for some of the processes mentioned above, and energy consumption. Also, this case is mainly related to the Climate Change and the water depletion impact categories. (d) Weighting process and final preparation: In refractory bricks production, the powdered material is weighted and prepared to press and convert into bricks. In the case of the refractory mass, the pressing process is not applicable. This process is related to climate change due to energy consumption requirements. (e) Drying and firing processes: These constitute the most energy-consuming processes because refractory materials must be dried and fired in special furnaces at very high temperatures (>1500 K) for at least 24 h. This process is strictly related to the climate change impact category due to energy consumption requirements. (f) Packaging, transportation, and final installation: These operations are strictly related to climate change, requiring energy and fuel consumption. (g) Waste disposal or recycling: The final step following the end of the LC includes two main activities: waste disposal and recycling or reusing. The decision to

Row materia l

Emmission to the air

Fig. 1.3 Life Circle for refractory products

Applying (Use)

Transportaon

End of Life Cycle (

Packaging

Row mterial Transportaion

Disposal

Drying and Firing

Preparaon (Crushing, Milling, sieving

Water Waste to the Water

Weighng, pereparaon for producon

Energy Dry Pressing (Forming)

Waste to the soil

Row Material Producons

10 A. Spyridakos et al.

1

The Multicriteria Features of Environmental Footprint. . .

11

handle the refractory waste depends on (1) the industrial machines and furnaces used, as this affects the degree to which the refractories have eroded and acquired impurities, (2) the composition and the quality of the refractories, as they have a different behaviour such as magnesium ones than clay ones. The deposition of spent refractory materials is also accompanied by relatively significant environmental impacts, while landfilling is required, mainly when used in industrial units that justify corrosion and admixture with components that give them permeable properties for the environment. Recycling them into new refractory materials or other materials (e.g. materials for road surfacing or construction materials) is the best case from the environmental point of view, but this can be done for spent materials that have no impact on the environment for the granulometry to be used. Therefore, the methodological approach for EF assessment based on EU L124 legislation (European Commission, 2013), which has similarities to the ISO standards 14040, 14025, and 14044, provides a structural framework through which the main factors of environmental impacts can be pictured quantitatively. As mentioned above, this approach is based on techniques where the inputs and emissions are calculated into equivalent measures, which are then normalised and weighted with parameters estimated under strong assumptions and utilising experts’ opinions. It is characterised as a multicriteria problem, and the proposed process for the EF Assessment is oriented to handle it in this way. The criteria in Multicriteria Decision Aid (MCDA) constitute monotone functions picturing the preferences variation across the values (qualitative or quantitative) for points of view governing the case examined. The set of criteria used in any case ought to satisfy the principles of consistency (Bouyssou, 1990; Roy, 1985, 1990), that is to say, for any set of Criteria G = {g1, g2, g3,. . .} and a set of alternative actions or situations (a1, a2, . . .}. The preference structure (P, I) is introduced where ai P aj when ai is preferred than aj and ai I aj when there is indifference between ai and aj. For every case examined, the set of criteria ought to satisfy the principles of consistency (exhaustivity, Cohesion, Redundancy), which can be summarised into the following: • Exhaustivity: For every criterion gi and a, b alternative actions or situations gi(a) = gi(b) = > aIb. In other words, the set of criteria is complete, and no criterion is missing. • Cohesion: For every pair of alternative actions or situations, if (a) exists at least one criterion gi where gi(a) > gi(b) and (b) for all the other criteria gj, gj(a) = gj(b), j ≠ i then aPb. This condition ensures that the used family of criteria respects the dominance. • Redundancy: The absent of one of the criteria violates the above mentioned properties of exhaustivity and Cohesion. In other words, the set of criteria used is complete. According to the consistency of the criteria used for EF assessment, some issues must be considered. The approaches for evaluating global EF either at the midpoint

12

A. Spyridakos et al.

or endpoint level are based on the calculation of the equivalent emissions and inputs coming from the LC of a production process or services offered. The equivalent emissions and inputs are assigned to the impact categories based on simple techniques with multiplications with predefined factors. The specific case is characterised by high complexity. One emission or input can influence more than one impact category, and more than one emission and input influence one impact category. The categories of impacts are not mutually independent. Also, arguments are posed concerning exhaustivity. Although the 16 impact categories are significantly important for the environment, it should be kept in mind that they are not the only ones. The research in this field is ongoing, and new research results lead to the enrichment and adaptation of impact categories. So, there is continuous progress in this field, and there is fluency while the scientific community is still working on a complete set of impact categories for EF Assessment. The assessment of Environmental Impacts proposed by ISO standards and used worldwide is based on processes assumed linear relationships, which concludes in the various stages of the process to multiplications with the identified factors for the purposes used. This comes into contraction with the fundamental relationships of environmental impacts to the quantities of emissions and inputs. The accurate relationship identification and modelling of the environmental impacts of emissions and inputs is a challenging process. On the one hand, the parameters’ accuracy cannot be checked efficiently; on the other hand, the linearity assumption does not represent a real-life situation. Therefore, in methodologies that measure environmental impacts, the relationship between cause and effect should be modelled in a way that is more consistent with reality. Another critical factor concerns the determination of the gravities of the individual environmental impacts in measuring the Environmental Footprint. Weightings have been determined by the EU after dialogues with experts and aggregating their opinion. The weighting vector is the result of a process that aggregates the importance of every environmental impact category and the uncertainty of the accuracy of the calculations in each of them in the method commonly used. This analysis raises many questions about the accuracy and consistency of the weighting vector derived from this analysis. Also, environmental category impacts are excluded from the study in case of high uncertainty as far as the weighting is concerned. The latter raises one more question concerning the consistency of the global EF model. The analyses proposed for the environmental footprint assessment should be examined from the starting stages, and a methodological framework should be determined that would incorporate the principles of Multicriteria Analysis to ensure the greatest possible consistency based on the limitations set by the Multicriteria Analysis. This task will require time and effort for studying to define the specific parameters of the problem and for confronting issues arising to achieve rational evaluation models satisfying the principles of the MCDA. The most capable theoretical trends for the EF Assessment can be based on Value System approaches of Multicriteria Analysis and, in particular, on Additive Value

1

The Multicriteria Features of Environmental Footprint. . .

13

System (Jacquet-Lagreze & Siskos, 1982; Keeney & Raiffa, 1976) described in the following formulas: Given a set of finite alternative actions A = {a1, a2, . . ., am} evaluated on n preference monotone criteria G = {g1,g2, . . ., gn}: n

U ð ak ð g Þ Þ = i=1

pi ui ðgi Þ, uðgi Þ = 0, u gi = 1, for i = 1, 2, ::, n and k = 1, 2, ::m n

pi = 1, pi ≥ 0, for i = 1, 2, ::, n i=1

Where ak(g) = (ak(g1), ak(g2), . . ., ak(gn)) is the evaluation vector of an alternative action ak on the n criteria, gi* and gi* are the least and most preferable levels of the criterion gi, respectively, and ui(gi), pi are the value function and the relative weight of the i-th criterion. Applying the Additive Utility Model requires adapting the process from the first steps of the Life Cycle Analysis and primarily focuses on categorisation and characterisation. The collection of the necessary data ought to be structured and adapted to satisfy the consistency principle of the criteria by providing the context of the correlations between emissions/inputs and the environmental impact categories as well as the degree or level of the relative consequences. One more fundamental issue that should be considered is the need for adaptations of the methodological framework for every category of products or service offered. Analysing a product or service’s life circle leads to measuring the impact, which usually concerns a subset of the 16 environmental impact categories identified at the categorisation. Therefore, using all 16 environmental impact categories is redundant, and it does not make sense to use or take all of them into the environmental footprint assessment. Worldwide, the development of rules for specific products category is followed, but this is a process under development. At the EU level, the Rules (Products Environmental FootPrint Products Category Rules) have been completed for 13 with wide acceptance by Collective and Scientific Bodies. Therefore, the methodological framework for the specialised formulation of the multicriteria approach in each product or service category can be pictured in the PEFRC, considering the criteria and the corresponding environmental impacts related to the complementary product or service. This ensures a general framework that allows comparability at the level of products and services categories and between the discrete products within each category. Regarding the case of Refractory Materials, as already analysed in the previous ones, the emissions and inputs are mainly focused on the category of climate change, material and secondarily, with much smaller impacts on the toxicity of water resources and the use of water resources. Therefore, the Environmental Impact Assessment should focus on these four of the 16 criteria.

14

1.4

A. Spyridakos et al.

Conclusions

The case of the Environmental Footprint assessment is strictly related to the multicriteria Paradigm. The environmental impacts are measured into predefined categories utilising specific standards discriminating the different points of view of the environmental impacts resulting from human activity. The EF assessment is of intense interest to the Scientific community, and there is much more to be studied in continuous development as it constantly evolves following the results of science and technology. According to the ISO 14025, 14044, and 14040 standards and the L124 directive of the EU, a methodological framework is followed based on the principles of multicriteria analysis. It proceeds with assessing the footprint indicators through approximate techniques and certain assumptions. In this research work, the multicriteria feature of EF was presented and analysed from the search for the appropriate methodological framework to measure the environmental impacts from the production and use of refractory as well as the handling of the waste. Also, issues concerning the EF assessment were identified whose further investigation might lead to the formulation of proposals for estimating the Environmental Footprint based on the principles and the structural components of Multicriteria Analysis. The application of the methodological approach for assessing the environmental footprint in refractory materials highlighted the requirements for further investigation of the process to provide a methodological tool that can be easily applied to support the comparability and interpretability of the outcomes. Finally, the implementation of an Additive Utility System is proposed, which can function in this case if the targeting from the first stages of the study of the Environmental Footprint is carried out based on the principles of consistency. Acknowledgement This research was supported by the European Regional Development Fund of the European Union and Greek national funds (Greek Secretariat for Research and Innovation– GSRI) through the Operational Program Competitiveness, Entrepreneurship, and Innovation (EPAnEK 2014–2020) under the call RESEARCH—CREATE—INNOVATE. Project: “Recycling of used refractories from various industries to produce alumino-silicate refractories, ceramics, and masses” (project code: T1EDK05442).

References Abubakar, I. R., & Bununu, Y. A. (2020). Low carbon city: Strategies and case studies. In W. Leal Filho, A. Marisa Azul, L. Brandli, P. Gökçin Özuyar, & T. Wall (Eds.), Sustainable cities and communities. Encyclopedia of the UN sustainable development goals (pp. 366–376). Springer International Publishing. https://doi.org/10.1007/978-3-319-95717-3_24 Arianpour, F., Kazemi, F., Golestani-Fard, F., & Rasti, M. (2007). Characterisation of spent MgO-C refractory bricks with emphasise on recycling. In September 2007 Aachen, Germany (pp. 598–601). Bortsie-Aryee, N., & Gabriel, C.-A. (2020). Ecological footprint: Pragmatic approach to understanding and building sustainable cities. In W. Leal Filho, A. Marisa Azul, L. Brandli, P. Gökçin Özuyar, & T. Wall (Eds.), Sustainable cities and communities. Encyclopedia of the UN

1

The Multicriteria Features of Environmental Footprint. . .

15

sustainable development goals (pp. 141–150). Springer International Publishing. https://doi.org/ 10.1007/978-3-319-95717-3_37 Bouyssou, D. (1990). Building criteria: A prerequisite for MCDA. In C. A. Bana e Costa (Ed.), Readings in multiple criteria decision aid (pp. 58–80). Springer. https://doi.org/10.1007/978-3642-75935-2_4 Bringezu, S., & Bleischwitz, R. (2009). Sustainable resource management: Global trends, visions and policies (1st ed.). Routledge. https://doi.org/10.4324/9781351279284 Čuček, L., Klemeš, J. J., & Kravanja, Z. (2012). A review of footprint analysis tools for monitoring impacts on sustainability. Journal of Cleaner Production., 34, 9–20. https://doi.org/10.1016/j. jclepro.2012.02.036 Deneen, M. A., & Gross, A. C. (2010). Refractory materials: The global market, the global industry. Business Economics, 45(4), 288–295. Energy Transitions Commission. (2017). Better energy, greater prosperity - Achievable pathways to low-carbon energy systems. https://www.energy-transitions.org/publications/better-energygreater-prosperity/. European Commission. (2013). Annex II: Product environmental footprint (PEF) guide in commission recommendation of 9 April 2013 on the use of common methods to measure and communicate the life cycle environmental performance of products and organisations (2013/ 179/EU). Official Journal of the European Union, L124(56), 6–106. https://doi.org/10.3000/ 19770677.L_2013.124.eng Fang, K., & Heijungs, R. (2015). Investigating the inventory and characterization aspects of footprinting methods: Lessons for the classification and integration of footprints. Journal of Cleaner Production, 108, 1028–1036. https://doi.org/10.1016/j.jclepro.2015.06.086 Fang, K., Heijungs, R., & de Snoo, G. (2013). The footprint family: Comparison and interaction of the ecological, energy, carbon and water footprints. Revue de Métallurgie, 110(1), 77–86. https://doi.org/10.1051/metal/2013051 Fang, K., Heijungs, R., & de Snoo, G. R. (2014). Theoretical exploration for the combination of the ecological, energy, carbon, and water footprints: Overview of a footprint family. Ecological Indicators, 36, 508–518. https://doi.org/10.1016/j.ecolind.2013.08.017 Fang, H., Smith, J. D., & Peaslee, K. D. (1999). Study of spent refractory waste recycling from metal manufacturers in Missouri. Resources, Conservation and Recycling, 25(2), 111–124. https://doi.org/10.1016/S0921-3449(98)00059-7 Fang, K., Song, S., Heijungs, R., de Groot, S., Dong, L., Song, J., & Wiloso, E. I. (2016). The footprint’s fingerprint: On the classification of the footprint family. Current Opinion in Environmental Sustainability., 23, 54–62. https://doi.org/10.1016/j.cosust.2016.12.002 Ferreira, G., López-Sabirón, A. M., Aranda, J., Mainar-Toledo, M. D., & Aranda-Usón, A. (2015). Environmental analysis for identifying challenges to recover used reinforced refractories in industrial furnaces. Journal of Cleaner Production, 88, 242–253. https://doi.org/10.1016/j. jclepro.2014.04.087 Hanagiri, S., Matsui, T., Shimpo, A., Aso, S., Inuzuka, T., Matsuda, T., Sakaki, S., & Nakagawa, H. (2008). Recent improvement of recycling technology for refractories. Special Issue on Refractory Technology in the Steel Industry. Nippon Steel Technical Report, 98, 93–98. Hoekstra, A. Y. (2009). Human appropriation of natural capital: A comparison of ecological footprint and water footprint analysis. Ecological Economics, 68(7), 1963–1974. https://doi. org/10.1016/j.ecolecon.2008.06.021 Horckmans, L., Nielsen, P., Dierckx, P., & Ducastel, A. (2019). Recycling of refractory bricks used in basic steelmaking: A review. Resources, Conservation and Recycling, 140, 297–304. https:// doi.org/10.1016/j.resconrec.2018.09.025 Huijbregts, M. A. J., Steinmann, Z. J. N., Elshout, P. M. F., Stam, G., Verones, F., Vieira, M., Zijp, M., Hollander, A., & van Zelm, R. (2017). ReCiPe2016: A harmonised life cycle impact assessment method at midpoint and endpoint level. The International Journal of Life Cycle Assessment, 22(2), 138–147. https://doi.org/10.1007/s11367-016-1246-y

16

A. Spyridakos et al.

Huppes, G., de Koning, A., Suh, S., Heijungs, R., van Oers, L., Nielsen, P., & Guinée, J. B. (2006). Environmental impacts of consumption in the European Union: High-resolution input-output tables with detailed environmental extensions. Journal of Industrial Ecology, 10(3), 129–146. https://doi.org/10.1162/jiec.2006.10.3.129 Jacquet-Lagreze, E., & Siskos, J. (1982). Assessing a set of additive utility functions for multicriteria decision-making, the UTA method. European Journal of Operational Research, 10(2), 151–164. https://doi.org/10.1016/0377-2217(82)90155-2 Joint Research Centre (European Commission), Cerutti, A. K., Pant, R., & Sala, S. (2018). Development of a weighting approach for the environmental footprint. LU, Publications Office of the European Union. https://data.europa.eu/doi/10.2760/945290 Keeney, R. L., & Raiffa, H. (1976). Decisions with multiple objectives: Preferences and value tradeoffs. John Wiley & Sons. Khan, A., Muhammad, F., Chenggang, Y., Hussain, J., Bano, S., & Khan, M. A. (2020). The impression of technological innovations and natural resources in energy-growth-environment nexus: A new look into BRICS economies. Science of The Total Environment., 727, 138265. https://doi.org/10.1016/j.scitotenv.2020.138265 Krausmann, F., Wiedenhofer, D., & Haberl, H. (2020). Growing stocks of buildings, infrastructures and machinery as key challenge for compliance with climate targets. Global Environmental Change, 61, 102034. https://doi.org/10.1016/j.gloenvcha.2020.102034 Kyriakopoulos, G. L., Kapsalis, V. C., Aravossis, K. G., Zamparas, M., & Mitsikas, A. (2019). Evaluating circular economy under a multi-parametric approach: A technological review. Sustainability, 11(21), 6139. https://doi.org/10.3390/su11216139 Matuštík, J., & Kočí, V. (2021). What is a footprint? A conceptual analysis of environmental footprint indicators. Journal of Cleaner Production, 285, 124833. https://doi.org/10.1016/j. jclepro.2020.124833 Monfreda, C., Wackernagel, M., & Deumling, D. (2004). Establishing national natural capital accounts based on detailed ecological footprint and biological capacity assessments. Land Use Policy, 21(3), 231–246. https://doi.org/10.1016/j.landusepol.2003.10.009 Poirier, J., Blond, E., de Bilbao, E., Michel, R., Coulon, A., Gillibert, J., Boussuge, M., Zhang, Y., Ryckelynk, D., Dusserre, G., Cutard, T., & Leplay, P. (2017). New advances in the laboratory characterization of refractories: Testing and modelling. Metallurgical Research & Technology, 114(6), 610. https://doi.org/10.1051/metal/2017068 Rees, W. E. (1999). The built environment and the ecosphere: A global perspective. Building Research & Information, 27(4–5), 206–220. https://doi.org/10.1080/096132199369336 Roy, B. (1985). Méthodologie multicritere d’Aide à la Decision. Economica. Roy, B. (1990). Decision-aid and decision-making. European Journal of Operational Research, 45(2), 324–331. https://doi.org/10.1016/0377-2217(90)90196-I Sharif, A., Baris-Tuzemen, O., Uzuner, G., Ozturk, I., & Sinha, A. (2020). Revisiting the role of renewable and non-renewable energy consumption on Turkey’s ecological footprint: Evidence from quantile ARDL approach. Sustainable Cities and Society, 57, 102138. https://doi.org/10. 1016/j.scs.2020.102138 Sinha, A., Mishra, S., Sharif, A., & Yarovaya, L. (2021). Does green financing help to improve environmental & social responsibility? Designing SDG framework through advanced quantile modelling. Journal of Environmental Management, 292, 112751. https://doi.org/10.1016/j. jenvman.2021.112751 Sinha, A., Sengupta, T., & Saha, T. (2020). Technology policy and environmental quality at crossroads: Designing SDG policies for select Asia Pacific countries. Technological Forecasting and Social Change, 161, 120317. https://doi.org/10.1016/j.techfore.2020.120317 Sinha, A., Shah, M. I., Sengupta, T., & Jiao, Z. (2020). Analyzing technology-emissions association in Top-10 polluted MENA countries: How to ascertain sustainable development by quantile modeling approach. Journal of Environmental Management, 267, 110602. https://doi.org/10. 1016/j.jenvman.2020.110602

1

The Multicriteria Features of Environmental Footprint. . .

17

Tan, E. C. D., & Lamers, P. (2021). Circular bioeconomy concepts—A perspective. Frontiers in Sustainability, 2, 701509. https://doi.org/10.3389/frsus.2021.701509 Tukker, A., & Jansen, B. (2006). Environmental impacts of products: A detailed review of studies. Journal of Industrial Ecology, 10(3), 159–182. https://doi.org/10.1162/jiec.2006.10.3.159 Van den Bergh, J. C. J. M., & Verbruggen, H. (1999). Spatial sustainability, trade and indicators: An evaluation of the ‘ecological footprint’. Ecological Economics, 29(1), 61–72. https://doi.org/ 10.1016/S0921-8009(99)00032-4 Vlachokostas, C., Achillas, C., Diamantis, V., Michailidou, A. V., Baginetas, K., & Aidonis, D. (2021). Supporting decision making to achieve circularity via a biodegradable waste-tobioenergy and compost facility. Journal of Environmental Management, 285, 112215. https:// doi.org/10.1016/j.jenvman.2021.112215 Wackernagel, M., & Rees, W. E. (1997). Perceptual and structural barriers to investing in natural capital: Economics from an ecological footprint perspective. Ecological Economics, 20(1), 3–24. https://doi.org/10.1016/S0921-8009(96)00077-8 Wackernagel, M., & Rees, W. E. (1998). Our ecological footprint: Reducing human impact on the earth. New Society Publishers. Wiedmann, T., & Barrett, J. (2010). A review of the ecological footprint indicator—Perceptions and methods. Sustainability, 2(6), 1645–1693. https://doi.org/10.3390/su2061645

Chapter 2

Fire Disaster Recovery and Resource Allocation Enabled by Firefighters’ Sustainable UAV Technology in Smart Cities Theodoros Anagnostopoulos and Yannis Psaromiligkos

Abstract Autonomous Unmanned Aerial Vehicle (UAV) technology is able to serve physical disaster recovery needs in Smart Cities (SC). Such critical situations require instant treatment to avoid scaling of the disaster in the whole area of a city. In case of a physical disaster such as a fire in the SC infrastructure is possible to lead to a high coverage area damage. Fire expansion may produce serious problems in disaster recovery and should be treated accordingly by certain municipality fightfighters which are dedicated to extinguish such fires. However, firefighters, vehicles, and UAVs that are available to treat a fire in the city are limited and cannot serve all the areas simultaneously thus transforming the upcoming situation to a resource allocation problem. In this paper, to handle such a problem there is not a need to serve each incident with the same priority since there are cases of low, medium, or high significance. Specifically, SC infrastructure has a limited number of UAVs to heal certain fire cases. On a trigger occurrence, such as a phone call from a fire section area, the nearest UAV reaches a certain fire incident where the assigned personnel at the SC control center assess the severity of the problem based on the video and audio provided by the camera and the microphone embedded in the UAV. Dedicated personnel then evaluate if it is a high significance fire incident and a vehicle with assigned firefighters is invoked to reach and fix the problem in real time.

2.1

Introduction

Planet earth is facing unexpected global warming conditions, which may cause mass scaling physical disasters. Citizens’ daily lives could suffer from emerging disasters in Smart City (SC) infrastructure. Such disasters may negatively affect citizens’ wellbeing in SCs green and sustainable ecosystems. Humans witness a plethora of T. Anagnostopoulos (✉) · Y. Psaromiligkos DigiT.DSS.Lab, Department of Business Administration, University of West Attica, Athens, Greece e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2023 A. Spyridakos (ed.), Multicriteria Decision Aid and Resource Management, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-34892-1_2

19

20

T. Anagnostopoulos and Y. Psaromiligkos

serious physical disasters in the last decades like fires, earthquakes, and floods. Citizens’ suffer from severe injuries or even loss of their lives due to these disasters. In this paper, we propose a control system architecture, which is based on Unmanned Aerial Vehicle (UAV) technology to treat the effects of a mass scale fire physical disaster to human population living in SCs. System’s architecture operation is assigned to the municipality headquarters control center where appropriate personnel is dedicated to orchestrate the SC recovery from certain fire disaster. Such a disaster consists of a severe security issue and should be faced as soon as possible by the SC municipality control center (Anagnostopoulos et al., 2021). A fleet of autonomous UAVs is located in several locations within the SC. Such fleet is available to provide remote evidence of a critical fire expansion in the SC’s buildings’ infrastructure. Specifically, the assigned municipality personnel are able to decide in real time a high significant fire incident and assign it to a dedicated fire vehicle with adequate firefighters to extinguish the fire from the affected location to prevent expansion in the greater SC area. Proposed control system architecture is able to extinguish certain fire incidents in the SC spatial area. However, fire recovery is considered as a resource allocation problem where a limited amount of firefighters, dedicated fire vehicles, and UAVs are constantly available for certain fires that may occur. Such municipality resources need to be invoked rationally to serve on real time certain fire incidents in a SC’s green and sustainable ecosystem. Certain use cases are getting through well-defined experiments to infer which is effective for large-scale SC adoption. Specifically, Use Case I (UC-I) is a centralized approach where UAVs are assumed to be located in the center of the SC, while another Use Case II (UC-II) assumes a decentralized distributed infrastructure to assess the proposed control system architecture efficiency. Concretely, current study adopts certain evaluation metrics, which are incorporated to infer the optimal use case for actual adoption by SCs. Structure of the paper is presented as follows. Section 2.2 examines related work in the research area. In Sect. 2.3 is presented the proposed control system architecture. Section 2.4 analyzes the proposed evaluation system parameters, while Sect. 2.5 describes certain metrics used to assess the adopted system. Section 2.6 defines the examined use cases. In Sect. 2.8 are performed the experiments. Section 2.9 discusses the observed results, while Sect. 2.9 concludes the paper and proposes future work.

2.2

Related Work

Firefighting of certain areas of a SC terrain is an interesting research area where researchers propose specific technical solutions to support firefighters in their activities to extinguish fire. Fire is a physical disaster. Concretely, fire can emerge simultaneously in certain areas of the SC infrastrucure. Such disaster may be faced by applying contemporary solutions, which can be adopted by the municipality control center to treat such a threat. A smart elevator-aided fire evacuation to face

2

Fire Disaster Recovery and Resource Allocation Enabled by Firefighters’. . .

21

an upcoming incident, which occurs in high-rise apartment buildings used by elderly and impaired citizens is proposed in (Fang et al., 2022). An inventive fire detection system, which is able to utilize Raspberry Pi for homes with advanced technology located in SCs is proposed in (Sheth et al., 2020). A study on urban fire risk and its assessment through an incorporated index system for SCs is proposed in (Sun, Lin, et al., 2019a). A robust smart solution for extensive fire safety in a large garage environment is proposed in (Antonov, 2019). Authors in (Yao et al., 2019) propose a contemporary fire evacuation system design for facing fire risk assessment for a SC’s stadium. Authors in (Song et al., 2020) analyze certain research factors toward fire extinguish behavior of a specific bubble atomizing nozzle used in metro modeling. In (Abuhamdah & Mehairi, 2015) research is performed in the area controlled by the Dubai Electricity and Water Authority (DEWA) focusing on a smart solution to mitigate the emerging substation fire incident in Dubai station. Researchers in (Acakpovi et al., 2021) propose the use of an innovative fire detection system, which is enabled with alarm utilities for supporting sustainable SC development. Research performed in (Mazur-Milecka et al., 2021) focuses on examining proactively a possible fire incident in SC area by incorporating an operational detection system, which uses thermal imaging technology. In (Sergi et al., 2020) authors propose an efficient fire management system, which is able to exploit the potentiality of Internet of Things (IoT) technologies as well as integrating the proposed solution in the context of Building Information Modeling (BIM) environments for SC wellbeing. A smoke flow analytical model, which is used under different exhaust mode in case of an emergency fire incident occurs in a subway carriage in SC infrastructure is proposed by researchers in (Gong & Zhou, 2019). Research effort in (Mahgoub et al., 2020) proposes an effective fire alarm system, which incorporates edge computing potentiality to provide a sustainable and green ecosystem for citizens’ wellbeing in SCs. An unobtrusive fire prevention system is proposed in (Zaher et al., 2021), which is able to monitor SC terrain for an upcoming fire incident. Such a system focuses on real time and online proactive operations to mitigate the fire risk in the city. An efficient system, which is based on a Regional Convolutional Neural Network (R-CNN) for exploiting video and fire sensing to support an indoor and outdoor unobtrusive surveillance system for smart SCs is proposed by authors in (Saponara et al., 2020). An extensive study is proposed in (Lin et al., 2016), which investigates the application of BIM technologies in cases of fire physical disasters. Such a system is able to handle emerging fire incidents in multiple location areas within the SC. Authors in research effort (Cheng & Zhang, 2019) perform a multiparametric study where they define experimentally the explosion safety distance, which should be followed in case of a Liquified Petroleum Gas (LPG) tank fire incident. In (He et al., 2015) it is proposed a road traffic fire system software, which is able to predict a fire incident influence in SC road networks. Such a system may be used to detour road traffic from alternative routes in the city to contribute to traffic balance thus avoiding bottlenecks in road junctions. Authors in (Zhang et al., 2021) propose a risk warning system to face an emerging fire hazard in SCs. Such system exploits the potentiality of big data and artificial

22

T. Anagnostopoulos and Y. Psaromiligkos

intelligence, which provide the framework where the proposed system is built and tested with real and synthetic data sources. An integrated proactive fire detection analytical system, which exploits IoT and image processing technology potentiality is proposed to serve upcoming smart city emergency fire incidents, as described in (Sharma et al., 2020). Researchers in (Walia et al., 2018) propose a dynamic pipeline technical approach to enable spatiotemporal fire risk early warning and prediction in the SC location area. A participatory design system, which is based on the dynamics of wearable computing technology is proposed by authors in (Klann, 2007). Such system enables firefighters with wearable devices, which helps them to achieve effective and on time fire extinguish operations when fighting with fire in SC’s buildings. In (Sun, Fang, et al., 2019b) authors propose a Bayesian Belief Network application, which is used to face SC fire risk. Such research effort incorporates historic data statistics and sensor data sources to feed the adopted model to operate online and in real time in case of an emerging fire incident. A domestic smart firefighting robot is proposed by researchers in (Ahmed et al., 2022), which incorporates multisensory fire detection mechanism along with a robust warning system built using Python Integrated Development Environment (IDE). Such firefighting robot is able to extinguish fire in dangerous areas that human firefighters cannot reach due to fire smoke and possible toxic air pollution, which is caused by an emerging fire incident. Research efforts analyzed in the contemporary literature focus on fire extinguish of a possible fire that may arise in a certain SC location. Specific system architectures along with supported technologies have been proposed to face such a physical disaster. However, analyzed research studies fail in facing the fire incident problem in great detail. Concretely, such approaches do not exploit, in many cases, the potentiality of remote assessment with a fleet of specific autonomous UAVs. Proposed research effort uses a resource allocation system architecture, which incorporates UAVs in the process of evaluating the significance of each fire incident as well to provide enough information to the municipality control center to engage adequate amount of firefighters and fire vehicles to extinguish the upcoming fire incidents in the SC area. Intuitively, observed incidents are prioritized as low, medium, or high significance emergency situations based on the assigned municipality personnel’s decision. After evaluating a high significance incident, a fire vehicle with a certain number of firefighters are invoked to treat the fire incident problem of the certain SC place. Current system architecture adopts specific use cases, one centralized and the other distributed, to face an emerging fire incident. Proposed use cases are validated and evaluated based on certain designed algorithms to infer the efficiency of the proposed system architecture approach.

2.3

Control System Architecture

Proposed system’s architecture incorporates specific detection and service algorithms, which are invoked to treat an emerging fire incident. These algorithms are able to mitigate the upcoming disaster risk as well as to provide recovery online and

2

Fire Disaster Recovery and Resource Allocation Enabled by Firefighters’. . .

23

Fig. 2.1 Control system architecture overview: (1) SC municipality control center, (2) fire incident, (3) UAV, (4) municipality fire vehicle, and (5) firefighter

in real time. Control system architecture overview is presented in Fig. 2.1. The proposed detection algorithm is input with the current status of the system, i.e., if there is a certain fire incident to serve, as well as the location of the incident in the SC’s infrastructure. Observed results of the model are the verification denoting the incident has been served by certain firefighter and fire vehicle resource allocation. In case a fire incident emerges, proposed system invokes the service algorithm providing the location of the incident in the SC area. The detection algorithm is presented in Table 2.1. In case the service algorithm is invoked, system architecture engages the nearest autonomous UAV to visit the fire incident location in the SCs infrastructure. Before commanding the round-trip, availability of the battery lifetime is checked, i.e., if it is greater than the estimated time to serve the incident, thus to verify that the selected UAV has enough energy to perform the task. If battery is drained, system engages subsequently the next nearest autonomous UAV to serve the upcoming fire incident location. This is a repeated dynamic process until an available UAV has adequate battery level. Selected UAV is navigated autonomously to the incident location by exploiting shortest path algorithm potentiality. On UAV arrival to the incident location the embedded camera and microphone are activated accordingly to provide the assigned personnel in the SC municipality control center an overview of the fire damage significance. If the incident’s significance is considered high by the assigned personnel, the system calls a fire vehicle to extinguish fire incident and recover the damage. Subsequently, when the incident has been served the status of the system model is updated to be ready to serve a consecutive fire incident. Service algorithm is presented in Table 2.2.

24

T. Anagnostopoulos and Y. Psaromiligkos

Table 2.1 Detection algorithm

# 1 2 3 4 5 6 9 10 11 12 13 14

Detection algorithm Input: Status, Loci //Status, incidence location Output: Status //Trigger status Begin Status = 0 //Originally there is no trigger to handle While (TRUE) Do If Status = 1 Then //If a phone call trigger happens Status ← Service(Status, Loci) //Invoke service algorithm to handle the event End If Return(Status) End While End

Table 2.2 Service algorithm

# 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

Service algorithm Input: Status, Loci //Status, incidence location Output: Status //Trigger status Begin Engage UAV Nearest to Loci While (TRUE) Do If (Battery Lifetime > Estimated Time) Then //If UAV has enough energy to perform task Navigate Autonomous UAV to Loci //Incorporate shortest path to reach Loci Enable Remote Diagnosis //Municipality personnel provide remote diagnosis If (Incidence Significance = High) Then Call a Fire Vehicle to Extinguish Fire //Transfer any injured citizen from Loci to hospital End If Status ← 0 //Indicate that incident has been served Return(Status) //Exit service algorithm Else Engage Another UAV Nearest to Loci End If End While End

2.4

Evaluation Parameters

Specific evaluation parameters are incorporated to assess the efficiency of the proposed fire incident use cases. Intuitively, the observed number of incidents occurred in such a physical disaster is decomposed to an average number of

2

Fire Disaster Recovery and Resource Allocation Enabled by Firefighters’. . .

25

Table 2.3 Evaluation parameters Parameter I L M H U B T D

Description Average total number of incidents Average number of low significance incidents Average number of medium significance incidents Average number of high significance incidents Average number of autonomous UAVs Average number of battery recharges Average amount of time required Average distance covered

incidents I, which are characterized by the assigned SC’s municipality control center personnel either as low, L, medium, M,or high, H, significance fire incidents. Concretely, the average number of autonomous UAVs used, U, to reach the problematic areas during average number of incidents is also considered by the system control architecture. UAVs operation is based on battery limitations. Specifically, battery needs recharge when drain in periodical time occurrence. In addition, an indicator of the average number of battery recharges is also proposed for experimentation, B, by the system. Subsequently, the average amount of time required, T, as well as average distance to be covered in each invocation, D, to serve a fire incident is equally important. Control system’s adopted evaluation parameters are presented in Table 2.3.

2.5

Evaluation Metrics

Computing the performance of the adopted system per certain use case can be achieved by defining specific evaluation metrics, which assess the adopted values of a given evaluation parameter. Assuming, that R1 = UI 2 ½0, 1 is the average number of UAVs per average incidence ratio, which is able to measure the average number of UAVs engaged per average number of incidents. Concretely, low value of R1 indicates an optimum use case since less average number of UAVs are used. H 2 ½0, 1 to be the average number of high Subsequently, let us assume, R2 = LþMþH significance incidences per average number of low, medium, or high incidents emerged. It holds that low value of R2 indicates an effective use case since less average number of high significant incidents occurred and need instant treatment. Let us, also, consider R3 = TB 2 ½0, 1 to be the ratio of average number of battery recharges per average amount of time required to serve an incident. In this case, a low value of R3 indicates an efficient way to handle average amount of incidents. Subsequently, assume, R4 = DT 2 ½0, 1 to be the ratio of average UAV velocity (i.e., ratio of average distance covered per average time required) observed during average UAVs certain call. A high value of R4 indicates optimum service average UAV velocity of the adopted system. Proposed evaluation metrics are presented in Table 2.4.

26

T. Anagnostopoulos and Y. Psaromiligkos

Table 2.4 Evaluation metrics Metric R1 =

U I

R2 =

H Lþ M þ H

R3 =

B T

2 ½0, 1

R4 =

D T

2 ½0, 1

2.6

2 ½0, 1 2 ½0, 1

Measures Average number of UAVs engaged per average total number of incidents Average number of high significance incidences per average number of low, medium, or high incidents occurred Average number of battery recharges per average amount of time required to serve an incident Average velocity (i.e., ratio of average distance covered per average time required) observed during average UAVs invocation

Use Cases

Adopted control system architecture is evaluated on specific categories of SC green and sustainable infrastructure. Let us define, UC-I, which is the case of a centralized system infrastructure where certain number of autonomous UAVs is located in a central base is station at the center of the SC spatial terrain. Concretely, assume that is incorporated a number N of UAVs to handle an emerging fire incident in the SC area. If a phone call occurred it triggers the control system, where the closest UAV is called to serve the upcoming fire incident. This is a dynamic and continuing process until all the available UAVs are incorporated. However, note that in this use case is assumed that the SC area is not divided ito specific area sectors, which in turn means that on a consecutive control system trigger a UAV is possible to engaged at any location within the whole SC spatial terrain. Subsequently, it can be inferred that in this case the system resources are not used efficiently and may not be sufficient to serve the upcoming citizens’ needs, online and in real time, to provide high-quality recovery services. UC-I is presented in Fig. 2.2.

Fig. 2.2 UC-I system overview

2

Fire Disaster Recovery and Resource Allocation Enabled by Firefighters’. . .

27

Fig. 2.3 UC-II system overview

Let us define another use case, namely UC-II, which focuses on a decentralized distributed infrastructure where the SC spatial terrain is divided into specific number of equal sized P sectors according to the number N of the available autonomous UAVs by the adopted SC control system architecture. In such a case, each sector has assigned a certain UAV to serve its fire recovery needs. It is assumed that there is a certain N number of base stations at the center of each SC sector. When a control system trigger emerges, it is invoked the nearest UAV, which is assigned to the closest sector. In this case, it also holds that the process is continued dynamically and repeatedly until all the available and charged UAVs are incorporated. Concretely, the main difference is that when a new trigger is occurred it is served by the nearest UAV, which is actually located in the closest neighbor sector of the emerging fire incident. UC-II is presented in Fig. 2.3.

2.7

Experiments

Certain experiments are performed in this research to assess the efficiency of each use case and define which is more effective in case of fire disasters in SCs. Control system architecture of the SC municipality is assumed to have N = 10 autonomous UAVs in the overall SC spatial area. In case of UC-I the autonomous UAVs are all located at the SC center, while in case of UC-II UAVs are equally distributed to the available P = 10 certain sectors of the SC terrain. Concretely, for UC-II it holds that for each sector is assigned a unique UAV. Experiments performed for a number of IT = 1000 iterations. For specific iteration, the control system architecture is called several times according to the incidents, IC, observed, which are following a random

28

T. Anagnostopoulos and Y. Psaromiligkos

Table 2.5 Experimental parameters

Parameter N P IT IC

Value 10 10 1000 (0,100]

distribution within the interval IC 2 (0,100] incidents. It is incorporated a random distribution of fire incidents in several locations of the SC to eliminate bias of the proposed system’s output results. Experimental parameters are presented in Table 2.5.

2.8

Results and Discussion

Proposed control system architecture is input with specific experimental parameters, which output the values as a result of certain experiments. Concretely, adopted research approach is based on the values of the evaluation parameters, which causes the system to result on certain metrics. These metrics are able to assess the efficiency of each use case. Subsequently, in Fig. 2.4 are presented the experimental results of R1 metric for the use cases UC-I and UC-II, respectively. It can be observed that R1 output values for UC-II are less than that of UC-I. Specific result values indicate that UC-II is an optimum use case since less average number of UAVs are used by the proposed control system architecture. Intuitively, output results for R2 metric are presented in Fig. 2.5. It can be denoted that values of R2 metric for UC-I are greater than that of UC-II. Specific outcome indicates that UC-II is an efficient use case since less average number of high significant incidents occurred and need treatment from the available resource allocation sustainable infrastructure.

Fig. 2.4 R1 metric observed values for UC-I and UC-II

2

Fire Disaster Recovery and Resource Allocation Enabled by Firefighters’. . .

29

Fig. 2.5 R2 metric observed values for UC-I and UC-II

Fig. 2.6 R3 metric observed values for UC-I and UC-II

Concretely, in Fig. 2.6 are presented the R3 metric results. However, note that R3 values for UC-II are less than that of UC-I, which is an indication that UC-II is a use case with an effective feature to serve average amount of fire incidents in the SC coverage terrain. Subsequently, R4 metric output results are presented in Fig. 2.7. It can be observed that values of r4 metric for UC-I are less numerically than that of UC-II, which is an indication that UC-II is a use case with optimum service of average UAV velocity for the proposed fire extinguish physical disaster recovery system architecture. In addition, a comparison of the observed results from the performed experiments in the current research effort leads to the inference that UC-II use case is optimum compared with the UC-I, since it has less R1, R2, and R3 output values than UC-I. Specifically, UC-II has higher R4 metric values than UC-I. The inferred efficiency is

30

T. Anagnostopoulos and Y. Psaromiligkos

Fig. 2.7 R4 metric observed values for UC-I and UC-II

based on the adopted evaluation metrics used to assess the effectiveness of both use cases as analyzed in Sect. 2.5, where is presented the evaluation methodology which leads to specific outcome of the current research in SC’s fire extinguish dynamic processes.

2.9

Conclusions and Future Work

Observed global climate warming effect leads to increase of physical disasters like fires, floods, and earthquakes. Such disasters affect citizens’ wellbeing since they cause severe injuries to the existing SCs’ population. It is proposed a control system architecture to face the effects of fire physical disasters to citizens living in green and sustainable SC ecosystems. A fleet of autonomous UAVs is incorporated, to serve fire incidents of certain significance, such as low, medium, and high emerged incidents. Adopted control system architecture is based on detection and service models to face an emergent fire incident after a physical disaster in SCs. The fleet of autonomous UAVs is located in specific areas within the SC spatial coverage terrain according to certain adopted use cases. Such use cases are divided into a centralized UC-I and a decentralized distributed UC-II, which are evaluated to assess to the effectiveness of the adopted control system architecture. Specific number of evaluation metrics and parameters are proposed to infer which use case is optimum for adoption by SCs in cases of fire physical disasters in the city’s infrastructure. This service is considered as a resource allocation problem where a limited number of UAVs, municipality fire vehicles, and firefighters dedicated personnel is assigned to treat an upcoming fire incident. In this research it is performed specific experiments where it is proved that UC-II is optimal compared with UC-I according to the adopted evaluation metrics. Future research directions focus on providing further exploitation of the adopted detection and service models

2

Fire Disaster Recovery and Resource Allocation Enabled by Firefighters’. . .

31

by incorporating distributed artificial intelligence technology and reinforcement approaches design. Such a kind of intelligent systems are able to serve the upcoming incidents based on stochastic and periodic knowledge of prior occurred fire incidents observed at the SC coverage terrain.

References Abuhamdah, I., & Mehairi, M. A. (2015). DEWA’s smart solution to mitigate the substation fire incident in Dubai station. In 1st IEEE international smart cities conference, Guadalajara, Mexico (pp. 1–5). Acakpovi, A., Ayitey, T., & Adjaliko, E. N. (2021). Innovative fire detection and alarm system for sustainable city development. In 1st IEEE international conference on cyber security and internet of things, France (pp. 30–36). Ahmed, S. M., Shiva, A. D. S., & Gunjan, V. K. (2022). Domestic smart fire fighting robot with multisensory fire detection and warning system using python IDE. In V. K. Gunjan & J. M. Zurada (Eds.), Proceedings of the 2nd international conference on recent trends in machine learning, IoT (Smart cities and applications. Lecture notes in networks and systems) (Vol. 237, pp. 47–55). Springer. Anagnostopoulos, T., Kostakos, P., Zaslavsky, A., Kantzavelou, I., Tsotsolas, N., Salmon, I., Morley, J., & Harle, R. (2021). Challenges and solutions of surveillance systems in IoT-enabled smart campus: A survey. IEEE Access, 9, 131926–131954. Antonov, S. (2019). Smart solution for fire safety in a large garage. In 1st IEEE international conference on creative business for smart and sustainable growth, Sandanski, Bulgaria (pp. 1–4). Cheng, X., & Zhang, Y. (2019). Study on explosion safety distance in LPG tank fire. In 2019 IEEE international conference on intelligent transportation, big data & smart city, Changsha, China (pp. 682–686). Fang, H., Qiu, H., Lin, P., Lo, S. M., & Lo, J. T. Y. (2022). Towards a smart elevator-aided fire evacuation scheme in high-rise apartment buildings for elderly. IEEE Access, 10, 90690–90705. Gong, C., & Zhou, R. (2019). Smoke flow analysis under different exhaust mode in case a fire occurs in a subway carriage. In 2019 IEEE international conference on conference on intelligent transportation, big data & smart city, Changsha, China (pp. 66–71). He, Y., Zhou, X., Du, S., & Ran, M. (2015). Traffic influence of road traffic based on VISSIM. In 2015 IEEE international conference on intelligent transportation, big data and smart city, Halong Bay, Vietnam (pp. 951–954). Klann, M. (2007). Playing with fire: Participatory design of wearable computing for fire fighters. In CHI EA ‘07 extended abstracts on human factors in computing systems, San Jose, USA (pp. 1665–1668). Lin, L. K., Hsu, Y. H., & Hsiao, Y. C. (2016). A study of applying BIM technique into fire disaster investigation system. In 8th IEEE international conference on measuring technology and mechatronics automation, Macau, China (pp. 32–35). Mahgoub, A., Tarrad, N., Elsherif, R., Ismail, L., & Al-Ali, A. (2020). Fire alarm system for smart cities using edge computing. In 1st IEEE international conference on informatics, IoT, and enabling technologies, Doha, Qatar (pp. 567–602). Mazur-Milecka, M., Glowacka, N., Kaczmarek, M., Bujnowski, A., Kaszynski, M., & Ruminski, J. (2021). Smart city and fire detection using thermal imaging. In 14th IEEE international conference on human system interaction, Gdansk, Poland (pp. 1–7). Saponara, S., Elhanashi, A., & Gagliardi, A. (2020). Exploiting R-C-NN for video smoke/fire sensing in antifire surveillance indoor and outdoor systems for smart cities. In 1st IEEE international conference on smart computing, Bologna, Italy (pp. 392–397).

32

T. Anagnostopoulos and Y. Psaromiligkos

Sergi, I., Malagnino, A., Rosito, R. C., Lacasa, V., Corallo, A., & Patrono, L. (2020). Integrating BIM and IoT technologies in innovative fire management systems. In 5th IEEE international conference on smart and sustainable technologies, Split, Croatia (pp. 1–5). Sharma, A., Singh, P. K., & Kumar, Y. (2020). An integrated fire detection system using IoT and image processing technique foe smart cities. Sustainable Cities and Society, 61, 102332–102354. Sheth, M., Trivedi, A., Suchak, K., Parmar, K., & Jetpariya, D. (2020). Inventive fire detection utilizing raspberry pi for new age home in smart cities. In 3rd IEEE international conference on smart systems and inventive technology, Tirunelveli, India (pp. 724–728). Song, X., Jian, C., & Jun-qing, Z. (2020). Research on fire extinguishing behavior of bubble atomizing nozzle in metro model. In 2020 IEEE international conference on intelligent transportation, big data & smart city, Vientiane, Laos (pp. 48–52). Sun, J., Lin, M., Wu, J., Liu, X., Sun, T., & Fang, H. (2019a). Study on urban fire assessment index system for smart cities. In 2nd IEEE international conference on electronics technology, Chengdu, China (pp. 609–613). Sun, J., Fang, H., Wu, J., Sun, T., & Liu, X. (2019b). Application of Bayesian belief networks for smart city fire risk assessment using history statistics and sensor data. In Data science. ICDS 2019. Communications in computer and information science (Vol. 1179, pp. 3–11). Springer. Walia, B. S., Hu, Q., Chen, J., Chen, F., Lee, J., Kuo, N., Narang, P., Batts, J., Arnold, G., & Madaio, M. (2018). A dynamic pipeline for spatio-temporal fire risk prediction. In Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining, London, United Kingdom (pp. 764–773). Yao, H. W., Zhang, D. S., & Liang, D. (2019). Fire evacuation design and fire risk assessment for a stadium. In 2019 IEEE international conference on intelligent transportation, big data & smart city, Changsha, China (pp. 650–653). Zaher, A., Al-Faqsh, A., Abdulredha, H., Al-Qudaihi, H., & Toaube, M. (2021). A fire prevention/ monitoring smart system. In 2nd IEEE international conference on smart cities automation & intelligent computing systems, Tangerang, Indonesia (pp. 31–36). Zhang, Y., Geng, P., Sivaparthipan, C. B., & Muthu, B. A. (2021). Big data and artificial intelligence based early risk warning system of fire hazard for smart cities. Sustainable Energy Technologies and Assessments, 45, 100986–100996.

Chapter 3

Resource Management: A Bi-Objective Methodological Approach for Routing in Crisis Situations Stamatios Vasalakis and Athanasios Spyridakos

Abstract This research work aims to describe a Resource Management (RM) approach which was developed in such a generalized way that allows its adaption and implementation in other high-stake sectors. There are a lot of cases that there are strong needs for immediate response and the best possible allocation of resources. The below methodological approach is characterized as a bi-objective dynamical and deterministic. The proposed methodological approach is based on the principles of Dynamical Programming (DP), which discretizes the optimization problems in stages and the techniques of Linear Programming (LP) to identify the optimal transportation plan at every stage utilizing live data from Google. This kind of problem is considered multiplicative because there are two principal objectives: (a) the emergency to transport the injured according to their triage scale and (b) the immediate movement of the patients to the hospitals. Also, the problem can be characterized as multiplicative because of (a) the existence of many start points and destination points of the ambulances with different distances and availabilities, (b) the destination points of one stage become the start points of the next stage, increasing the multiplicity of the problem, and (c) the need to achieve global optimization of the time response in relation to the emergency and priorities of the patients’ condition. The above proposed methodological approach is illustrated through a case study involving the timely transfer of injured people as a result of a natural disaster to the nearest hospitals at a local level.

S. Vasalakis (✉) · A. Spyridakos Laboratory of Business Informatics and Quantitative Methods, Department of Business Administration, School of Administrative, Economics and Social Sciences, University of West Attica, Athens, Greece e-mail: [email protected] © Springer Nature Switzerland AG 2023 A. Spyridakos (ed.), Multicriteria Decision Aid and Resource Management, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-34892-1_3

33

34

3.1

S. Vasalakis and A. Spyridakos

Introduction

The speed of reaction that patients or injured ones will be given care by the medical staff in organized medical structures is in many cases crucial for the successful confronting of the case. In natural disasters with human victims, the high number of injured requires the direct activation and the exploitation of the available resources in order to confront the increased requirement for immediate medical support. The effective management of the available resources constitutes essential factor for the effective treatment of these undesirable situations. The availability of information is the key, giving decision-makers insights of the required resources, their availability and the level of direct utilization. Instead of randomly allocating resources, it is useful to rely on an advance resource management plan so as to avoid delays and consequently to save many of our fellow human beings. The utilization of RM practices supports the effective planning, scheduling, and allocating people, vehicles, costs, etc., in order to have better results and to reduce the undesirable impacts. Thus, the case of handling that transportation of patients or injured in situation of natural disasters includes two main dimensions. The first one is related to the management of ambulances so as to minimize the response time, and the second one is resulted by the hospital resources limitations and constrains, such as the availability to treat injured in emergency healthcare departments, the availability of hospitality, and the availability of the medicine staff. Also, prioritization of the cases (injured or patients) has to be taken into consideration so as to cure the urgent ones prior to others for whom there is no need for immediate treatment. Especially, in natural disasters that have far-reaching effects on human lives, all available ambulances have to be utilized so as to arrive to the accident area as soon as possible and transfer the injured ones to the medical facilities (hospitals) which can provide comprehensive medical services, which are critical for life and for the course of their health development. This constitutes a logistic problem with special conditions and parameters which have to be taken into consideration that differentiate it from a traditional logistic problem increasing the complexity. Generally, on several transportation logistic problems, predefined transport routes are followed, having precise and updated data required in order to estimate a routing plan, minimizing the cost or time, exploiting the available people, vehicles, and/or infrastructures. The complexity of above-mentioned case is analyzed in the following. The ambulances usually are based at different locations (Hospitals or the Ambulance Stations) or are engaged in other cases. So, the location of the available ambulance which will be used for the injured or patients’ transportation affects the response time. The available ambulances usually are limited in relation to the number of the injured collections. So, some or all the ambulances have to make more than one routes for the collection of all the injured. Each destination point for an ambulance is the starting point of the next route. In addition, the time taken by the ambulance to transport patients includes a high level of uncertainty, while factors such as the traffic and the conditions of the local road network change from time to

3

Resource Management: A Bi-Objective Methodological Approach for Routing. . .

35

Fig. 3.1 Illustrated problem

time. Also, it has to be taken into consideration that the second route of an ambulance is affected by the choice of the first route. Another significant parameter is the health condition of the injured and the prioritization to be given to every one of them. The provision of medical care starts from the moment the rescuer arrives at the scene of the accident, but the substantive care would be provided in the medical facilities (hospitals, healthcare institutions, etc.). People who are seriously injured take precedence over others who are slightly injured (Hodge et al., 2013). The availability of the hospital resources (human and infrastructures) is limited in each hospital, especially at regional levels. So, the best allocation of resources supports the confronting the consequences of a natural disaster such as an earthquake or a flood or a fire or an epidemic or a road accident (Euth, 2000) (Fig. 3.1). In accidents of high severity, is required the immediate response only in one destination and manage the ambulance’s fleet so as to provide medical care solutions as soon as possible. Also, restrictions are posed while hospitals and ambulance fleet have to handle other not massive situations of the daily routine. A specific number of injured can be transported and served on every turn by the ambulance fleet and a specific number of injured can be treat on every turn by the hospitals. Also, a limited number of injured can receive the immediate medical care due to hospitals’ resource restrictions (humans and infrastructures). On the below diagram the problem is illustrated. In traditional logistic case studies, multiple and conflicting objectives are faced. While, there are restrictions on the supply and the demand. The transportation objects have to be transported from m sources to n destination points and the capacities of the sources are a1, a2, ..., am and the demands of destinations are b1, b2, ..., bn, respectively. The costs (which can be the transportation fees, the time, the

36

S. Vasalakis and A. Spyridakos

Fig. 3.2 Representation of multiple objectives in the transportation problem

damage cost, cost of damage, or safety of delivery, etc.) of transporting a unit from source i to destination j is notated Cij. A variable Xij represents the unknown quantity to be shipped from source i to destination j. The objectives are to minimize the total cost of transportation, delivery time, and/or damage cost. Let Z1, Z2, ..., Zk be k objectives that have to be minimized. It is always assumed that the balance condition holds (i.e., that the total demand is equal to the total supply) (Nomani et al., 2017) (Fig. 3.2). With these assumptions, the Multi-Objective Linear Transportation Problem (MOLTP) can be formulated as follows: m

n

Min Z k X ij = i=1 j=1

C kij X ij , k = 1, 2, 3, . . . , K

Subject to m

X ij = ai , j = 1, 2, 3, . . . , m i=1 n

X ij = bi , i = 1, 2, 3, . . . , n j=1

X ij ≥ 0 i = 1, 2, 3, . . . , m and j = 1, 2, 3, . . . , n Simple linear programming practices cannot be suggested because the environment is unstable and the existence of high degree of uncertainty and the decision taken in one round of the ambulance fleet mobility affects the next one. The goals are to maximize the service quality and simultaneously to minimize the fleet time cost. So, the case requires the identification of a set of routes and schedules that balance conflicting objectives, which are (a) the emergency to support/transport the injured according to their triage scale and (b) the immediate movement of the patients to the hospitals (Beaudry et al., 2010). The nature of this case study leads this research work to the exploitation of the features of dynamic programming because (a) there is a better confrontation of the uncertainty and the liquidity related to the involved parameters and (b) provides the methodological tool to handle the mobility of ambulance at one round taken into consideration the results of the previous round, seeking for optimization of the cost. The main contribution of this research work is the proposition of a methodological framework which can be exploited in real-world cases under the above-mentioned circumstances (uncertainty, complexity, dynamic environment, crucial to human being) It is worth to be mentioned that dynamic programming methodological frames have not been utilized extensively in this kind of case studies yet.

3

Resource Management: A Bi-Objective Methodological Approach for Routing. . .

37

The paper consists of the introduction and four more sections. In the second section, a literature review is presented concerning relative previous research works. Following, the proposed methodological framework is analyzed in detail and is illustrated through a case study in the fourth section. Finally, in the last section, the conclusions of this research work are summarized and the new perspectives coming from the results of this activity are outlined.

3.2

Literature Review

The identification of an optimal solution to the classical transportation problem requires the determination of the objects to be transported from each start point to specific destination/terminal points, satisfying the conditions coming from the sources’ availability and the destination’s demands so that the total transportation cost to be the minimum. The available resources and the demands on transportation problems usually are not deterministic, including low or high degree of uncertainty. This is the reason why the stochastic modeling is required in order to face the issues coming from uncertainty (Cooper, 1978; Mahapatra et al., 2013). Consequently, on the one hand, stochastic modeling constitutes a main stream to handle the uncertainty into the factors, and, on the other hand, requires more in-depth statistical and computational ability instead of the simple deterministic models. Moreover, there are a lot of research works that confront this kind of problem by a classical heuristics or metaheuristics procedure. A heuristic, or heuristic technique, is any approach to problem-solving that uses a practical method or various shortcuts in order to produce solutions that may not be optimal but are sufficient given a limited timeframe or deadline (Zhao et al., 2015). Also, a metaheuristic is a high-level problem-independent algorithmic framework that provides a set of guidelines or strategies to develop heuristic optimization algorithms (Sörensen et al., 2018). A heuristic solution method was also provided, since the supply and the demand cannot be estimated precisely but can be identified with some given probability distributions (Amirteimoori, 2011; Bertsimas et al., 2019). In such cases, a transportation plan with maximum relative efficiency is of major importance (Cordeau, 2006). However, these techniques require considerable workload for the analysis and review the stochastic parameters/factors, as well as the relative assumptions without checking the utility or the criticality of its factor which is not worth to follow this path in some cases. Moreover, a significant framework that we can solve this kind of problem is the tabu search algorithm. It was created by Fred W. Glover in 1986 and formalized in 1989. The tabu search takes a potential solution to a problem and checks its immediate neighbors in the hope of finding an improved solution. The tabu search is an optimization technique that uses a guided local search procedure that avoids local optimization and rejects moves to points already visited in the search space by means of the so-called tabu list (Beaudry et al., 2010; Hanne et al., 2009; Miyazaki et al., 2012).

38

S. Vasalakis and A. Spyridakos

Hanne et al. (2009) and Beaudry et al. (2010) exploited a tabu search algorithm putting additional constraints and priorities (urgent vs. normal) in order to manage risk situations which added complexity, especially for intra-hospital transportation. In the case of non-emergency transport of patients to/from hospitals, the vehicle can be reconfigured to provide staff seats, patient seats, stretchers, and wheelchairs. Nevertheless, the transfer of patients to the same hospitals that some patients have already gone is mandatory due to the liquidity of the parameters such as the capacity of every hospital and the traffic congestion. Furthermore, (Parragh, 2011; Parragh et al., 2012; Schilde et al., 2011; Schilde et al., 2014) a lot of research works in this field add more restrictions such as the capacity, the driver-vehicle assignments, the maximum shift lengths, and the mandatory driver breaks. A real-world application is coming from Hong Kong Hospital Authority (HKHA) where memetic algorithms were utilized in order to face needs of emergency response in cases of high risk. Memetic algorithms (Tripathy et al., 2018) constitute extensions or adaptations of the traditional genetic algorithms which are based on natural functions/observation such as ant’s colonies algorithm and bee’s colonies algorithm (Dorigo & Gambardella, 1997; Pham et al., 2005). Usually, memetic algorithms use a local search technique to reduce the likelihood of the premature convergence. The proposed by HKHA method, the ambulances have to be disinfected between consecutive trips so as to evade the spread of disease, as well as a critical dilemma is emerged while there can be patients which are not served due to the expected bad potentiality according to a decision that it is taken by the responsible medical staff (Zhang et al., 2015). Nonetheless, there is not always certainty for the progress of the human health, especially in situations where the decisions have to be taken in limited time. Detti et al. (2017) proposed a methodological framework where the patients could choose the transport provider and a tabu search algorithm was proposed in order to optimize the responds factors. In very difficult situations, patient’s preferences are of minor importance and it is significant to transport all the patients to the closest hospitals on time so as to save as many lives as possible. Some research works are based on assumptions concerning user requests which considerate are known in advance so as the vehicle routes to be organized ahead of time and initials solutions are given utilizing dynamic programming approaches (Desrosiers et al., 1986; Psaraftis, 1980; Psaraftis, 1983). Madsen et al. (1995) improve the dynamic programming approach by the utilization of insertion algorithms (one or two-phases) so as to be applicable in real-world situations. Insertion or online algorithm is a sorting algorithm in which the elements are transferred one at a time to the right position (Madsen et al., 1995). Coslovich et al. (2006) comments for the above-mentioned approach are focused to the fact that there is a need of a few hours in a static problem, while dynamic programming is much faster responding. Moreover, an online algorithm has been introduced recently which there is a signal communication scheme between the trip dispatcher and the algorithm is used to improve utilization of the available idle time (Lois & Ziliaskopoulos, 2017). On our case, in real-life applications, the most difficult thing is to determine the exact time for the transportation which is needed by the ambulances to transport the

3

Resource Management: A Bi-Objective Methodological Approach for Routing. . .

39

patients to the nearest hospitals due to the traffic or other parameters. In this paper, we assume that the supply, demand, and cost per unit values are exactly known. In addition, we utilize a backward recursion of dynamic programming and multicriteria decision analysis. Last but not least, the problem has been identified as multi-objective linear programming problem.

3.3

Methodology

Dynamic Programming is based on the recursive of a multistage decision processes (Bellman, 1957) and it is used for the confrontation of complex optimization problems where it is not effective to solve it utilizing linear programming (LP) or nonlinear programming (NLP) techniques. Since 1957, there were a lot of research works which utilizing dynamic programming approach for the effective confronts of complex optimization problems. The backbone of the proposed methodology is based on the Bellman’s original broad conception, which exploits the recursion features dividing the whole problem into stages so as to handle the complexity and simultaneously the results of a stage feed the next one. The deterministic multistage decision process is characterized by an “initial state” x1 and functions ft and Lt as follows: The function ft determine relations between “controls” {ut} and “states” {xt} according to xtþ1 = f t ðxt , ut Þ

1≤t≤N

The states xt’s are presumed to be real n-tuples and the controls ut’s are m dimensional. The set of decision stages, i.e., the domain for the index t are the first N positive integers, or the set of all positive integers for an infinite horizon process. Function ft(xt,ut) will be referred to as the “dynamics,” and Lt(xt,ut) as the “singlestage loss” function. With any sequence u = {ut} of controls (such a sequence being known as a “policy”), one associates a real-valued objective function defined by N

Lt ðxt , ut Þ

K ð uÞ = t=1

The goal is to construct a policy “u*” which minimizes the objective function K(u). The “feasible” controls are those which satisfy a vector-valued state-stage dependent constraint of the form gt ðxt , ut Þ ≤ 0

1≤t≤N

The procedure for the finite horizon optimal control problem begins by recursively solving the functional equation (Bellman, 1957) VN(x), VN-1(x), . . ., V1(x) with

40

S. Vasalakis and A. Spyridakos

V t ðxÞ = min u ½Lt ðx, uÞ þ V tþ1 ðf t ðx, uÞÞ t = N, N - 1, . . . , 1 Where VN+1(x)  0. The minimization is with respect to controls satisfying. Let ut(x) denotes a control minimizing the bracketed term in the above equation Vt(x). It is readily proven that the policy “u*” determined by x1* = x1 and ut = ut xt , xtþ1 = f t xt , ut

t = 1, . . . , N

Is a global minimizer of K(u); “u*” is a solution of the optimal control problem (Yakowitz, 1982). The complexity of our problem, which consists of the above-mentioned characteristics, leads to the adoption of a methodological approach which ensures that a decision at one stage will affect the results of a decision at the next stages, exploiting all the above-mentioned Bellman’s concept so as all the portions of an optimal trajectory to be optimal trajectories themselves. Thus, the stages are considered interdependent each other and each stage includes corresponding sub-stages. Generally, in dynamic programming, it is easier to resolve an optimization problem when the problem is breaking down in to a set of sub-problems. Also, the output of a sub-stage is utilized triggering the process for the assessment of the optimal solution of the next stage (Rardin, 2022). So, by the assessment of the optimal solutions of all the sub-problems, we can conclude to an optimal solution by tracing of the trajectory. In other words, the solution can refer to the estimation of the optimal combination of successive decisions. The relation between the value of the whole problem and the values of the sub-problems is used to be called the Bellman equation. There are two alternative approaches for handling the recursivity between the successive stages: (a) forward recursion and (b) backward recursion. Forward recursion chain follows a direction moving from the first stage to the last stage for the assessment of the optimal solution while backward recursion chain functions in an opposite way (Bettinger et al., 2017). Without limitation of generality and for the scopes of the representation of the proposed approach the backward recursion approach will be adopted. So, the following steps are implemented: (a) the problem is divided into distinct stages; (b) we start the research for the optimal solution for the last stage; (c) in every stage and for every alternative state the process is focused to the search of the optimal way to reach the state from the previous stage. This process continues until to reach the first states. The optimal solution(s) is achieved by tracing the results of every state and stage in a forward path way aggregating the ft results (Rardin, 2022). The below diagram presents the workflow of dynamic programing methodology (Fig. 3.3). For the purposes of the case examined in this research work, an adapted and enriched backward chaining technique of dynamic programming will be used. The states for us will be the outcome of every ambulance, taking into account the chosen patient, hospital, and route that it will be selected. The stage for us will be every separate round and the controls will be our restrictions. We consider that we are at the state x1 when everything begins and we will construct a policy “u*” which

3

Resource Management: A Bi-Objective Methodological Approach for Routing. . .

41

Fig. 3.3 The workflow of dynamic programming methodology

minimizes the objective function. The examined case in this research work is more complex since (a) the severity of each patient’s injury should be determined and (b) transport and treatment time should also be determined. As per the first one, the selection of each patient is based on the classification of Australasian Triage Scale (ATS) (Hodge et al., 2013). As per the second one, time depends on two factors: the travel time and the time it will take to receive health services taking into account the availability of hospital staff and infrastructure which means that not all patients can go to the same hospital because that’s how we will have dead lock. In addition, there will be some sub-stages where we will find where each ambulance should go to nearest hospital if it is chosen, taking into account the time cost of each one. The composition of these sub-stages will give us the state for each stage. After each stage, the time of each ambulance which is needed so as to transport a patient to available hospital will have to be recalculated taking into consideration that the previous states will affect the new ones. Τhe inputs we have at each sub-stage are the ambulance times to the nearest available hospitals and the outputs will be the final options/solutions/states of the ambulances to each available hospital for each sub-stage. More analytical, on each sub-stage, calculation concerning every ambulance will be taken place regarding the time cost for every available hospital and after that the selection of the faster route will be decided. In each sub-stage, each ambulance can be moved separately because each one is at a different starting point. So, at the beginning, a check for each ambulance which is the preferable route has to be taken, by comparing the times required for each hospital. After that, a centralized table will be created. On the basis of the aggregate table, we choose the ambulances that will move taking into account the restrictions we have. The selected patients will be as many as can be transferred to the nearest hospital, as long as they do not exceed the allowed number that can be transported in that round in total. Also, another significant restriction that we cannot neglect is the fact that we will not exceed the number of patients on the accident area. In addition, patients with the most serious injuries will be the ones to be transported and on the next sub-stages

42

S. Vasalakis and A. Spyridakos

the rest patients. After that, the selected ambulances and the hospital have to be removed so as to repeat this process with subsequent hospitals until one of our restrictions is not satisfied. It is self-explanatory, that if the restriction that is no longer satisfied is the fact that there are patients at the accident area, then our problem ends. In addition, the time consumed by the selected ambulances of the previous stage should also be added. So, the final time cost will be equal with the previous time cost plus the necessary time cost to be made in this round. Also, the final number of patients who were served by the hospitals will be equal with the previous final number of patients who were served by the hospitals plus the necessary change to be made in this round. On every next stage, the above procedure has to be repeated until the determination of the overall route result. The below diagram presents the workflow of backward recursion technique (Fig. 3.4). Also, taking decisions based on multiple different criteria with help from the Multiple Criteria Decision Analysis (MCDA) tool can then make things clear (Janse, 2018). Thus, we utilize the multi-criteria decision analysis so as we can minimize the time-cost which is needed by the ambulances to reach to the nearest hospitals taking into account all relevant constrains such as the number of patients which is served on every turn by the Hospitals and the maximum number of patients which will be served by the Hospitals. Last but not least, it should be noted that the terminal point reached by an ambulance is a starting point in the next turn. The below diagram presents the movement of ambulances (Fig. 3.5). First of all, we have a natural disaster during which we have some people who need an emergency medical attention. Secondly, we have to classify the patients based on the Australasian Triage Scale (ATS) which is our first objective (Hodge et al., 2013). Also, we have to calculate transport times to the hospitals using Live Google Data Times which the ambulances need so as they can take the patients from the accident area to the nearest hospital. Before we make the transport, we have to check if there are people who need an emergency medical attention except the first time that we are aware of this and check the availability of hospital rooms taking into consideration that each hospital can serve a certain number of people per round. At this time, we utilize the backward recursion technique so as to calculate the destination of each ambulance and in what order they will leave the waiting queue. After that, a specific number of ambulances can move to the accident area and pick up a certain number of patients. Furthermore, the next step of our methodology is to make the transportation to the nearest available hospitals and drop off the patients, utilizing the solutions of the backward technique of dynamic programming. At the end we have to check if all injured people transported to hospitals. If not, we have to determine from the beginning the new Live Google Data Times for the patients in order to transport them to hospitals taking into account that the new time is equal with the previous final time cost plus the necessary change to be made in this round. Also, the final number of patients who were cared for by the hospitals is equal with the previous final number of patients who were served by the hospitals plus the necessary change to be made in this round. In terms of mathematical formulation of our methodology it is supposed that there are some stages and some sub-stages of the problem which are referred to the

Resource Management: A Bi-Objective Methodological Approach for Routing. . .

Fig. 3.4 The workflow of Backward Recursion Technique

3

43

44

S. Vasalakis and A. Spyridakos

Fig. 3.5 The movement of ambulances

hospitals (n). It is assumed that the quantities available (ambulances) at the starting points, demands at the destination points (hospitals), and the time cost per unit of the quantity are precisely given. In addition, it is assumed that Sn will be the number of patients who can be served by the hospitals (n). Also, Xn will be the number of patients who were served by the hospitals (n). Last but not least Cn(Xn) will be the time cost of transportation of Xn patient to every hospital (n). So, the retroactive relation that is valid is the below.

F n ðSn Þ = min fðCn ðX n Þ þ F nþ1 ðSn - X n Þg X n ≥ 1, X n ≤ Sn With the objective of Min T =

Subject to the constraints:

a n=1

ðC n ÞðX n Þ where (n = 1, 2, , a) (hospitals).

3

Resource Management: A Bi-Objective Methodological Approach for Routing. . . a

ðCn ÞðX n Þ ≤ a Xn ≥ 0 ðIntegersÞ n=1

a n=1

Xn ≤

a

S n=1 n

a n=1

45

Xn = 5

In terms of restrictions the total cost has to be less than or equal to the total cost of every movement of the ambulances. The number of patients who will be served must be equal to zero or more than zero and this number must be integer. The number of patients who were served should be less than or equal to the number of patients who can be served. The number of patients who were served on every stage must be equal to 5. In addition to all the above, in each new stage, the final time cost is equal with the previous final time cost plus the necessary change to be made in this round. Last but not least, the final number of patients who were served by the hospitals is equal with the previous final number of patients who were served by the destinations plus the necessary change to be made in this round. a n=1

=

ðC n ÞðX n Þ = a

S n=1 n-1

þ

a n=1 a

ðC n - 1 ÞðX n - 1 Þ þ

a n=1

ðCn ÞðX n Þ

a

S n=1 n

S n=1 n

Another significant issue that under no circumstances should we neglect is the fact that if we change a small parameter like the available transportation on the nearest hospital and we increase this number, the total solution will be changed dramatically and we will diminish the time cost. On the other hand, if we diminish this number, we will have a worst scenario and the total time cost will be increased by causing catastrophic consequences for the patient’s health because we have to transport them to remote hospitals and we will increase the possibility for someone to die. That’s why we have to calculate everything from the beginning when we have a change on our parameters. Last but not least, we create an indicator which showed us the relationship between the treatment acuity and the transportation time which each ambulance needs. When this indicator is higher of 1, the patient is more likely to recover. When this indicator is between 0.5 and 1, the patient is more likely to be transferred to an upper level on this scale. When this indicator is lower than 0.5, the patient is almost certain to change scale, and when this indicator is closer to zero (0), the patient has high possibilities to lose his life (Vasalakis & Spyridakos, 2022).

3.4

Illustration

In order to determine the optimal number of patients that should be transported from the accident area to different near hospitals for obtaining the minimum transportation time cost, the following information were collected by utilizing live data coming from available e-platforms.

46

S. Vasalakis and A. Spyridakos

Fig. 3.6 Operation of software

Special software developed in order to extract live data and calculate the trip time between two different places. The addresses of the scene of the accident, ambulances, and hospitals must be entered in this software. A calculation concerning the transportation time cost has been taken place, which takes into account the traffic and the fastest available routes (Chi Brander Inc, 2017). There are also many parameters that can be selected regarding the distance type (meters, minutes, hours, or kilometers). The below diagram presents the operation of the above-mentioned software (Fig. 3.6). For the scopes of this illustration the case study examines (6) six start point and (3) terminal point (Hospitals). At the beginning, there are (2) ambulances on each start point. Moreover, we made an assumption that we have only 15 patients who are classified based on the Australasian Triage Scale (ATS) (Hodge et al., 2013). Also, we have to make (3) three stages and every stage will have (3) three sub-stages in order to transport everyone to the nearest hospitals because we assume that we can transport only five patients per turn. In Table 3.1, the maximum times of each ambulance have to make so as to transport the corresponding patient are presented (Hodge et al., 2013). Nevertheless, the time costs of per start point to the terminal point are given in Table 3.2. Every Hospital has limited available beds (Sn) and it can serve a specific number of patients in every turn. Nikaia Hospital can serve 8 patients in total and 2 patients in every turn. Panagia Odigitria (P.O.) hospital can serve 4 patients in total and 1 patient in every turn. Metropolitan hospital can serve 14 patients in total and 3 patients in every turn. Stage 1 Backward recursion of dynamic programming will be utilized on every ambulance separately. On the first sub-stage (A 1/N) a check will be taken place so as to decide on which hospital the ambulance can transport a patient. As the below procedure for the first ambulance, the selected hospital is the number one (Fig. 3.7). The same procedure will be followed for all the ambulances so as to secure which is the nearest hospital. As far as we can see, Panagia Odigitria hospital is the nearest

3

Resource Management: A Bi-Objective Methodological Approach for Routing. . .

47

Table 3.1 Classification of patients based on Australasian Triage Scale (ATS) Patient No 3 14 7 8 9 15 10 11 1 4 12 2 5 6 13

Degree of severity of injury (1—minor injury, 5— severe injury) 5 5 4 4 4 4 3 3 2 2 2 1 1 1 1

Treatment acuity (max wait time) 0 min 0 min 10 min 10 min 10 min 10 min 30 min 30 min 60 min 60 min 60 min 120 min 120 min 120 min 120 min

hospital, and we can select the ambulance number 3 because we can select only one ambulance for this hospital. So, we can select the first one patient who can be served by the hospital Panagia Odigitria. After that, we can remove Panagia Odigitria hospital and the time cost of the selected ambulance. Thus, the table of the ambulance transport time will be changed as below (Table 3.3). On the second sub-stage (A 2/N) a check will be taken place so as to see if we transport all the allowed patients for this stage. If not, a new check will be taken place in order to decide on which hospital the ambulance can transport a patient. As the below procedure for the ambulance number 1, the selected hospital is Nikaia hospital (Fig. 3.8). The same procedure will be followed for all the ambulances so as to secure which is the nearest hospital. As far as we can see, Nikaia hospital is the nearest hospital, and we can select the ambulance number 4 and number 11 because we can select only two ambulances for this hospital. So, we can select the second two patients who can be served by the Nikaia hospital. After that, we can remove Nikaia hospital and the time costs of the two selected ambulances. Thus, the table of the ambulance transport time will be changed as below (Table 3.4). On the third sub-stage (A N/N) a check will be taken place so as to see if we transport all the allowed patients for this stage. If not, and because we don’t have any other hospital, we will put every ambulance time cost on an ascend order and we will select the remaining available and allowed patients who can be served by this hospital taking into account that we have to select the minimum time costs. On

Hospitals

Νikaia P.O. Metropolitan

Ambulances 1 2 44.68 44.68 39.85 39.85 48.15 48.15

Table 3.2 Ambulance transport times 3 40.1 35.27 43.57

4 40.1 35.27 43.57

5 47.78 42.95 51.25

6 47.78 42.95 51.25

7 41.7 36.87 45.17

8 41.7 36.87 45.17

9 44.73 39.9 48.2

10 44.73 39.9 48.2

11 40.75 35.92 44.22

12 40.75 35.92 44.22

48 S. Vasalakis and A. Spyridakos

3

Resource Management: A Bi-Objective Methodological Approach for Routing. . .

49

Fig. 3.7 Backward recursion of Ambulance number 1

our case, we have already selected the five available patients who can be served by the hospitals (Table 3.5). So, at this time a check will be taken place so as to see if we transport all patients to the hospitals, but we have to proceed on the next stage because we have ten more patients to transport on the nearest hospitals. Stage 2 On second stage, all the previous transportation times will be changed. The new time cost is equal with the previous final time cost plus the necessary change to be made so as the ambulances go on the destination point (Hospital), utilizing new live data from Google maps Function. So, on the first sub-stage of the second stage (B 1/N), the time costs of ambulances to the hospitals are given on Table 3.6. The same procedure as the first stage will be followed and the next five patients will be transported by the ambulances according to Table 3.7. All constraints are already satisfied and (1) one ambulance will go to hospital Panagia Odigitria, (2) two ambulances will go to Nikaia hospital, and (2) two ambulances will go to Metropolitan hospital. So, at this time a check will be taken place so as to see if we transport all patients to the hospitals, but we have to proceed on the next stage because we have the last five patients to transport on the nearest hospitals. Stage 3 On third stage, all the previous transportation times will be changed. The new time cost is equal with the previous final time cost plus the necessary change to be made so as the ambulances go on the destination point (Hospital), utilizing new live data from Google maps Function. So, on the first sub-stage of the third stage (K 1/N), the time costs of ambulances to the hospitals are given on Table 3.8. The same procedure as the first stage will be followed and the last five patients will be transported by the ambulances according to Table 3.9. All constraints are already satisfied and (1) one ambulance will go to Panagia Odigitria hospital, (2) two ambulances will go to Nikaia hospital, and (2) two ambulances will go to Metropolitan hospital.

Hospitals

Νikaia Metropolitan

Ambulances 1 2 44.68 44.68 48.15 48.15

Table 3.3 Ambulance transport times 4 40.1 43.57

5 47.78 51.25

6 47.78 51.25

7 41.7 45.17

8 41.7 45.17

9 44.73 48.2

10 44.73 48.2

11 40.75 44.22

12 40.75 44.22

50 S. Vasalakis and A. Spyridakos

3

Resource Management: A Bi-Objective Methodological Approach for Routing. . .

0

S2X2

Number of Patients who were served 0 1

1

51

F2(S2)

0

0+0=0

1

0+48,15=48,15 39,85+0=39,85 39,85

Nikaia Hospital (S12 month, 10 °C); cold (more than 2 months of ice cover, mean temperature of warmest month 100 km2; 800 words) have outlier characteristics and were therefore, excluded. The filters were mostly applied during the formation of the positive corpus (D+). Text operations were identical to both corpuses and included: lowering the letters of the words, removal of stop words, punctuations, and numbers, creation of a custom dictionary with insignificant words (i.e., Athens, restaurant), elimination of rare words of the corpus, etc. Table 8.2 summarizes some key characteristics of the two corpuses. In both, the number of restaurants sampled is close to 12% of the total available restaurants. The comments (documents) included in D+ is significant higher due to the skewed distribution of rating. Finally, the unique words (vocabulary) refer to the number of words that topic modeling methods utilized, and the increased number found in D- is the result of the lengthier and more detailed comments when a negative experience is reported.

8.7.2

LDA Setting

This section addresses the creation and comparison of different LDA models and issues regarding the LDA setting, namely the selection of the number of topics and the hyper-parameter (α,β) setting. An important LDA parameter is the number of topics (k), which should be decided based on the specific problem and by utilizing prior knowledge. In our study, we did not have enough information to set an appropriate level for k and we decided to create multiple LDA models for a different number of topics. Therefore, we created a series of LDA models starting from the minimum number, which is k = 2, up to a sufficient number, which was set as k = 60. For every different level of k we used Gibbs sampling for determining the posterior probability of the latent variables. An efficient number of iterations was used (1500). We set the initial values

160

D. Novas et al.

Fig. 8.5 Selecting number of topics using probabilistic coherence

of hyper-parameters aligned to Griffiths and Steyvers (2004) propositions and updated every ten iterations based on the values θd. What makes a good topic is difficult to pinpoint precisely and various metrics have been proposed to address the issue, such as perplexity (Blei et al., 2003) and probabilistic coherence metrics (e.g., UCI by Newman et al. (2010) and UMass by Mimno et al. (2011)). Perplexity, in our dataset, was found to be unstable and we turn our attention towards coherence metrics, where we found probabilistic topic modeling as the most suitable. This measure examines a set of the top M most relevant words per topic (we set M = 10) and between two words of the topic (wi,wj) the amount of P(wi|wj)—P(wi) is calculated, where wj is more probable (frequent) than wi within the topic. The domain of values has been defined in [-1,1], where all the differences between the top M words are averaged and consequently the probabilistic topic’s coherence. Then, by averaging all available topic coherences, it was possible to have a single measure at the LDA model level, which corresponded to the k parameter. The suggestion is to select the k-th model that maximizes the probabilistic topic coherence. The next figure (Fig. 8.5) illustrates the aforementioned process and each dot on the lines represents an individual LDA model, trained at the corresponding corpus. The results suggested a low level of probabilistic coherence. Taking into account the underlying calculation method, two major causes related to the corpus vocabulary could be drawn. Firstly, the vocabulary associated with each corpus is considerable large, thus a significant number of words have low probability (frequency). Secondly, a significant part of the words are highly correlated (e.g., diner-night) but they also statistical independent due to the semantics. High correlation might be falsely caused when a word is part of a restaurant’s name (e.g., Acropolis tavern) and/or the proportion of comments among restaurants. To this end, we found that for the positive corpus (D+) an adequate selection was 36 topics and regarding the negative corpus (D-) 24 topics. For setting LDA we followed the propositions of Wallach et al. (2009). The authors suggested that the best LDA setting implies an asymmetric α Dirichlet prior for documents over topics and a symmetric β Dirichlet prior for topics over words. On the one hand, the α Dirichlet prior was calculated using an optimization scheme during the Gibbs sampling iterations and on the other hand we set β = 0.05 for all the

8

User Comments as a Resource to Rank with Multiple Criteria: The Case. . .

161

Table 8.3 Indicative topics and the α Dirichlet prior Corpus Positive corpus (k = 36)

a 0.218

Topic Topic 1

0.152

Topic 4 Topic 33 Topic 12

0.027 Negative corpus (k = 24)

0.241

0.132

Topic 17

0.033

Topic 11

Top five terms Found, reviews, tripadvisor, decided, meal Wine, service, experience, dining, excellent Service, price, quality, excellent, top Service, wait, time, table, minutes Reviews, tripadvisor, average, service, disappointed Hard, card, rock, pay, credit

Topic Interpretation The decision has been based on existing positive reviews in TripAdvisor, which is very common within such platforms A diner with nice wine, among others, offered a very nice experience. A valuable restaurant selection because it combines different dinning aspects Long waiting time for a table. Such issue is very frequent especially during the summer holidays. Tripadvisor’s reviews mislead the expectations and the tourists express disappointment Problems with the payment system of the restaurant

words. Profoundly, the α Dirichlet prior expresses the researcher’s belief about how a topic is distributed among the document and the higher the value the more probable is a document to contain a topic. The next table (Table 8.3) illustrates three topics per corpus, the asymmetric value of a, the related topic followed by the top five most frequent terms and a brief explanation by an industry expert. Table 8.3 topics were selected based on asymmetric α Dirichlet prior different levels of values (highest, median, and minimum). For example, regarding the positive corpus, Topic 1 has the highest α value, Topic 4 is close to the median, etc. The asymmetric value of α also found to have a linear relationship with the topic prevalence measure, so at a generic level it is possible to argue that a high α value indicates how much of a comment is associated with a topic. The values of α in the negative corpus are lower on average compared to positive corpus. Through reviewing the data, we concluded that the negative corpus has qualitative difference compared to the positive corpus and in more detail the former is more precise and factual, while the positive corpus tends to express overall experience and perceptions with generic words. It should also be noted that Topic 1 in the Positive Corpus reveals that the readers value highly TripAdvisor’s (in this case positive) reviews but the same also stands for negative reviews which was evident in our Topic 17 in the negative corpus.

8.7.3

Ranking All Restaurants

In this section, we propose a multi-criteria mechanism that ranks restaurants based on the comments and the ratings provided by the users. Before delving into the details, we shortly discuss the transformation of the LDA results to a manageable

162

D. Novas et al.

Fig. 8.6 Average assignment probability of topics over restaurants (Rest. and Top. refers to restaurant and topic accordingly)

format from the multi-criteria method, as it is the interface between two research fields. An important random variable of LDA is the per-document topic proportions (θd). Considering that a collection of documents refers to a restaurant, it is possible to group LDA’s assignments at a restaurant level, through the capitalization of a simple descriptive function (e.g., median) for the assignment’s distribution. Figure 8.6 exhibits such transformation and shows the topics assignment for 14 anonymized restaurants for both corpuses. Figure 8.6, reviewed either by columns or rows, provides some initial findings. In more detail, reading by columns it is possible to identify which topics are frequent (colored tiles of the heat map) and which are both frequent and intense (darker color). Similarly, viewing the heatmap by rows, an overview is acquired regarding the mixture of positive–negative aspects per different restaurant. For example, Rest.44 has a mixture of both positive and negative reviews and it seems having a significant problem regarding Topic 1 (in Negative corpus) but it also excels in Topic 18. In the contrary, Rest.281 has not any negative assignments and it seems that most of the positive comments are discussed in Topic 19. By managing topics as criteria, it is feasible to switch from the topic modeling area to multi-criteria decision-making and consequently create an efficient ranking mechanism. Assuming that each topic is a different criterion, the LDA exports the matrix θd fþ, - g from which derive the tables cn,i containing the average probability of the appearance of a topic in the comments of a restaurant. The average probability is the membership function μT i ðxÞ of each topic. For each restaurant, the membership fþ, - g functions of its topics occupy a row in matrix cn,i . However, some multi-criteria methods require use of the criteria weights as well. In our study, the weights are the ½ normalized values of the topics prevalence wi 2 cd ,  = þ , - and the subsets (T1, T2,. . .,Tm) are fuzzy variables (topics). As an example let us consider the negative comments for two restaurants to distinguish superiority of the one over the other. Table 8.4 depicts the weights of each topic. Table 8.5 contains the values of negative topics’ membership functions for restaurants 1 and 2 (Rest.1, 2). The

8

User Comments as a Resource to Rank with Multiple Criteria: The Case. . .

163

Table 8.4 Negative topics weights w

T1 0.04

T2 0.02

T3 0.07

T4 0.04

T5 0.02

T6 0.09

T7 0.02

T8 0.05

T9 0.02 f-g

Table 8.5 Average assignment of the n-th restaurant over the i-th topic cn,i Rest. 1 Rest. 2

T1 0.01 0.07

T2 0.01 0.02

T3 0.04 0.06

T4 0.04 0.03

T5 0.03 0.01

T6 0.15 0.12

T7 0.06 0.04

T8 0.08 0.01

T9 0.01 0.03

T10 0.02

... ...

T24 0.05

(negative comments) T10 0.01 0.03

... ... ...

T24 0.02 0.02

columns of this matrix are the topics’ membership functions. In other words, the intersection of the 3-rd row and 9-th column is the value of the membership function for the 9-th topic in the 2-nd restaurant and equals to 0.03. For example, we calculate where Rest.1 is superior to Rest.2 in Table 8.5 as follows: s=

wi , where I = f1 ≤ i ≤ 24 : Rest:1 is superior to Rest:2 in ith topicg = 0:51 i2I

I = f4, 5, 6, 7, 8,11,13,15,16,17,18,20,22,24g Similarly, we proceed with calculation on all pairs of alternatives to create relevant s. Then, all these s for each restaurant are summed based on function: Sn ½ -  =

si , where si is the superiority of nth restaurant over all the rest i≠n

This way we have the negative ranking. Relatively, we work to calculate the positive ranking. Finally, we subtract the obtained values of the negative from the positive rankings, over λ, to obtain the final ranking: Rn = ð1 - λÞ  Sþ n - λ  Sn

where λ is the rate that we want negative comments to be considered over the positive ones. If λ is equal to 0.5, the positive and negative comments are distributed evenly. Table 8.6 depicts how rating changes based on λ using our data. We calculated multiple λ, where for λ equal to 0.1, 10% weight of the negative comments over the positive ones is considered, etc., and included the average star ranking of the corresponding restaurant from TripAdvisor as well as the total stars per restaurant in two groups: low stars (1–3 stars) and high stars (4–5 stars). The results indicate

164

D. Novas et al.

Table 8.6 Partial ranking results of our method for the top rankings for each λ and some of the last in the ranking

Restaurant Rest.42 Rest.228 Rest.11 Rest.81 Rest.94 Rest.54 Rest.281 Rest.282 Rest.284 Rest.280 Rest.289

Star ranking mechanism Average High Low stars stars stars ranking (4 5) (1 2 3) 18 363 43 76 2089 44 43 261 229 67 448 208 2 49 51 1 39 39 0 33 1 0 68 3 0 42 4 0 60 10 0 30 5

Proposed ranking mechanism

λ = 0.1 1 2 3 4 23 34 146 157 182 185 195

λ = 0.2 1 2 5 11 3 4 45 63 97 101 120

λ = 0.3 12 14 31 32 3 4 1 2 5 8 9

λ = 0.4 24 29 59 54 9 10 1 2 3 4 5

λ = 0.5 44 50 101 92 15 16 1 2 3 4 5

that a restaurant ranks higher than the average star ranking for λ = 0.1 (e.g., 42, 228, 11, 81). However, an issue arises whether it is better to consider isobaric the negative and positive comments (λ equal to 0.5, 0.4) or, since the negative comments are considerably less, to take the negative comments proportionally to the positive ones (λ equal to 0.1, 0.2, 0.3). It is evident from Fig. 8.6 and Table 8.6 that for λ = 0.5 restaurants with minimal or not at all negative comments precede others. As observed from Table 8.6, there is a divergence of the proposed mechanism over the benchmark one, when λ = 0.1, λ = 0.2. In other words, there is a divergence when the negative comments are considered proportionally to the positive ones whereas going towards isobaric consideration of negative and positive comments the results from both mechanisms converge.

8.8

Discussion

In the big data era, the mass collected information becomes meaningful through proper processing. Such a processing is ranking mechanisms which are a new research area and it is evident that there are issues which need to be addressed (Rindova et al., 2018). It is not known how current ranking mechanisms work exactly since they are not published and therefore it is not known if they are efficient (Orlikowski & Scott, 2014). This gap is very stressful for management since the lack of evidence implies inability to make the right decisions. In the case of restaurants there are even less researches due to the complexity that arise from the subjective and

8

User Comments as a Resource to Rank with Multiple Criteria: The Case. . .

165

variable nature of the users’ comments. In this study, the comments of restaurants’ users in TripAdvisor were the resources used for multi-criteria decision-making. TripAdvisor’s popularity ranking mechanism materializes quality (rates), recency (latest ratings), and quantity of comments. The proposed methodology adopts recency as an important factor that identifies properly the current situation. However, there were issues that emerged considering quality and quantity of comments. In regard to quality, we encountered an occasional divergence among the rating provided by a user and its corresponding comment. In regard to quantity, it was observed that restaurants with a large number of comments (some negative among many positive) were ranked, according to TripAdvisor, higher than a restaurants that had overall fewer comments but they were all positive. In other words, excellent restaurants that do not have a significant mass of comments are undermined in the shuffle. Furthermore, since lesser quality restaurants are suggested higher, more customers visit them and consequently more comments are generated. The exponential distribution of comments resembles the Matthew effect (Rigney, 2010) where a mass amount of comments leads to further advantage (additional comments) and restaurants with a relative small amount of comments strive to overcome the commentary shortage. As a result, a vicious cycle is created that cancels the significance of quantity. Brought together quality and quantity mishaps in TripAdvisor’s ranking mechanism result in a deceiving ranking and users highlight that, when they comment that the proposed ranking does not correspond to reality. On the other hand, it should be noted that other users praise TripAdvisor’s ranking. Actually, both negative and positive dispositions towards TripAdvisor appear equally in this study’s topics and reveal that there are some issues that derive partially from the stars’ rating and more from comments’ volume. In the proposed ranking mechanism any volume of comments is acceptable as long as the sample is statistically significant and other than that the actual volume does not affect the ranking. Instead, all negative and positive comments are analyzed to extract the ranking of the restaurants. Through the research it became evident that λ is a factor that determines the ranking. Specifically, the issue is that users tend to complain but still give three or more stars which in turn classifies the comment as positive. When λ is given a lot of weight (considering λ equal to 0.4 and 0.5) the ranking we derive through the comments is closer to the average star ranking. Contrariwise, when the negative comments are considered proportionally to the positive ones (λ equal to 0.1, 0.2, 0.3) the results deviate from the average star ranking. This rises a dilemma whether or not to consider the negative comments isobaric to the positive ones or take them proportionally. Since, when taking negative and positive comments isobarically converges towards stars ratings this study proposes to weight positive and negative comments equally and regardless of the comments’ volume. In other words, we propose a ranking based on the λ which is between 0.4 and 0.5 and we deviate from considering more significant the large volume of the positive comments but we handle them as if they have equal weight with the fewer negative comments. Otherwise, the negative comments are not surfaced which leads to comments such as “selection of place was due to very

166

D. Novas et al.

high ratings from TripAdvisor . . .the place is awful ...I recommend to avoid this place.” After all, it is the negative comments that contribute to improvements. The proposed ranking mechanism is derived solely from the users’ comments and significantly differs from TripAdvisor’s Popularity Ranking. In the former, comments were handled at a document level with a qualitatively manner in order to produce topics and as a consequence set up the performance criteria, while in the latter the quantitative aspect (volume) of comments is considered. Another difference is related to the participation of the rating scale. In our case, the rates were only used to split the corpus into positive and negative comments. Contrary, TripAdvisor’s Popularity Ranking claims that rating along with recency and comments’ quantity are the important factors that constitute their own ranking approach. The utilization of LDA in the context of tourist industry was found valuable because it transforms unstructured data (comments) into a valuable resource that supplies the ranking mechanism. In our case, the higher probabilistic coherence of the negative corpus (D-) indicates that complaints tend to be more precise and factual (e.g., “we waited 30 minutes for a table”), compared to the positive experiences where the use of generic words describing perceptions might exist across different topics (e.g., “we had a lovely night”). Therefore, we suggest topics extracted from the negative comments as a source of pitfalls to avoid/corrective actions to undertake and the positive topics as a confirmatory basis of what tourists are looking for. Regarding LDA’s results, the high yielded number of topics urges a thorough study that will identify and remove insignificant topics for the ranking process. We acknowledge that this work has some limitations. As earlier discussed, we consider the proposed mechanism as preliminary, thus we recognize the incorporation of a sliding time window that weighs the most recent comments as more significant/accurate, which is an important characteristic to enhance the validity of the results. Moreover, the validation of the study is based on a benchmark mechanism which assumes that most of the platforms examine the quantitative (star rating) rather than the qualitative (comments) aspect of ranking, as this work suggests. The adopted benchmark rating scheme suggests the lack of a ranking ground truth employed by the platforms, therefore we consider the generalization of the findings as limited.

8.9

Conclusions

We have proposed a ranking mechanism based on the qualitative characteristics of tourists’ comments. We have argued that the qualitative aspects of comments are more instructive compared to the quantitative scale rating. The proposed mechanism is based on LDA to create the evaluation dimensional space, and the ranking follows the principles of MCDM by utilizing a λ parameter. We examined the proposed model for sample restaurants using the TripAdvisor platform and evaluated the results compared to a simple quantitative star rating scheme. We suggest that future

8

User Comments as a Resource to Rank with Multiple Criteria: The Case. . .

167

works will deal with alternative ranking mechanisms utilizing the qualitative aspects of user-generated content. Regarding the future improvements of the proposed method, the options are few: study and evaluate n-gram language models, examine fuzzy topic modeling, apply the method for hotels, attractions, etc.

References Behzadian, M., Kazemzadeh, R. B., Albadvi, A., & Aghdasi, M. (2010). PROMETHEE: A comprehensive literature review on methodologies and applications. European Journal of Operational Research, 200(1), 198–215. Bilgihan, A., Seo, S., & Choi, J. (2018). Identifying restaurant satisfiers and dissatisfiers: Suggestions from online reviews. Journal of Hospitality Marketing & Management, 27(5), 601–625. Blei, D. M. (2012). Probabilistic topic models. Communications of the ACM, 55(4), 77–84. Blei, D., & Lafferty, J. (2006). Correlated topic models. Advances in Neural Information Processing Systems, 18, 147. Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent Dirichlet allocation. Journal of Machine Learning Research, 3(Jan), 993–1022. Botti, L., & Peypoch, N. (2013). Multi-criteria ELECTRE method and destination competitiveness. Tourism Management Perspectives, 6, 108–113. Botti, L., Petit, S., & Zhang, L. (2020). Strategic decision concerning tourist origins portfolio: A decision process based on the ELECTRE method and applied to French Polynesia. Tourism Economics, 26(5), 830–843. Brandt, T., Bendler, J., & Neumann, D. (2017). Social media analytics and value creation in urban smart tourism ecosystems. Information & Management, 54(6), 703–713. Burgess, S., Sellitto, C., Cox, C., & Buultjens, J. (2009, June). User-generated content (UGC) in tourism: Benefits and concerns of online consumers. In ECIS (pp. 417–429). Büschken, J., & Allenby, G. M. (2016). Sentence-based text analysis for customer reviews. Marketing Science, 35(6), 953–975. Cheong, H. J., & Morrison, M. A. (2008). Consumers’ reliance on product information and recommendations found in UGC. Journal of Interactive Advertising, 8(2), 38–49. Choo, E. U., Schoner, B., & Wedley, W. C. (1999). Interpretation of criteria weights in multicriteria decision making. Computers & Industrial Engineering, 37(3), 527–541. Corrente, S., Greco, S., & Słowiński, R. (2013). Multiple criteria hierarchy process with ELECTRE and PROMETHEE. Omega, 41(5), 820–846. Daneshvar Rouyendegh, B., & Erol, S. (2012). Selecting the best project using the fuzzy ELECTRE method. Mathematical Problems in Engineering, 2012. Deerwester, S., Dumais, S. T., Furnas, G. W., Landauer, T. K., & Harshman, R. (1990). Indexing by latent semantic analysis. Journal of the American Society for Information Science, 41(6), 391–407. Dubois, D. J. (1980). Fuzzy sets and systems: Theory and applications (Vol. 144). Academic Press. Feldman, R. (2013). Techniques and applications for sentiment analysis. Communications of the ACM, 56(4), 82–89. Filieri, R., & McLeay, F. (2014). E-WOM and accommodation: An analysis of the factors that influence travelers’ adoption of information from online reviews. Journal of Travel Research, 53(1), 44–57. Gan, Q., Ferns, B. H., Yu, Y., & Jin, L. (2017). A text mining and multidimensional sentiment analysis of online restaurant reviews. Journal of Quality Assurance in Hospitality & Tourism, 18(4), 465–492. Gandibleux, X. (Ed.). (2006). Multiple criteria optimization: State of the art annotated bibliographic surveys (Vol. 52). Springer.

168

D. Novas et al.

Ganzaroli, A., De Noni, I., & van Baalen, P. (2017). Vicious advice: Analyzing the impact of TripAdvisor on the quality of restaurants as part of the cultural heritage of Venice. Tourism Management, 61, 501–510. Griffiths, T. L., & Steyvers, M. (2004). Finding scientific topics. Proceedings of the National Academy of Sciences, 101(suppl 1), 5228–5235. Guo, Y., Barnes, S. J., & Jia, Q. (2017). Mining meaning from online ratings and reviews: Tourist satisfaction analysis using latent Dirichlet allocation. Tourism Management, 59, 467–483. Ho, W., Xu, X., & Dey, P. K. (2010). Multi-criteria decision making approaches for supplier evaluation and selection: A literature review. European Journal of Operational Research, 202(1), 16–24. Hofmann, T. (2001). Unsupervised learning by probabilistic latent semantic analysis. Machine Learning, 42(1–2), 177–196. Iwata, T., Saito, K., Ueda, N., Stromsten, S., Griffiths, T. L., & Tenenbaum, J. B. (2005). Parametric embedding for class visualization. In Advances in neural information processing systems (pp. 617–624). Jacquet-Lagreze, E., & Siskos, J. (1982). Assessing a set of additive utility functions for multicriteria decision-making, the UTA method. European Journal of Operational Research, 10(2), 151–164. Jacquet-Lagreze, E., & Siskos, Y. (2001). Preference disaggregation: 20 years of MCDA experience. European Journal of Operational Research, 130(2), 233–245. Jamshidi, S., Rejaie, R., & Li, J. (2019). Characterizing the dynamics and evolution of incentivized online reviews on Amazon. Social Network Analysis and Mining, 9(1), 22. Jeacle, I., & Carter, C. (2011). In TripAdvisor we trust: Rankings, calculative regimes and abstract systems. Accounting, Organizations and Society, 36(4–5), 293–309. Jia, S. (2019). Measuring tourists’ meal experience by mining online user generated content about restaurants. Scandinavian Journal of Hospitality and Tourism, 19(4–5), 371–389. Keeney, R. L., & Raiffa, H. (1993). Decisions with multiple objectives: Preferences and value trade-offs. Cambridge University Press. Kirilenko, A. P., Stepchenkova, S. O., Kim, H., & Li, X. (2018). Automated sentiment analysis in tourism: Comparison of approaches. Journal of Travel Research, 57(8), 1012–1025. Krippendorff, K. (2018). Content analysis: An introduction to its methodology (4th ed.). SAGE. Laksono, R. A., Sungkono, K. R., Sarno, R., & Wahyuni, C. S. (2019, July). Sentiment analysis of restaurant customer reviews on TripAdvisor using Naïve bayes. In 2019 12th international conference on information & communication technology and system (ICTS) (pp. 49–54). IEEE. Lei, S., & Law, R. (2015). Content analysis of TripAdvisor reviews on restaurants: A case study of Macau. Journal of Tourism, 16(1). Limberger, P. F., Dos Anjos, F. A., de Souza Meira, J. V., & dos Anjos, S. J. G. (2014). Satisfaction in hospitality on TripAdvisor.com: An analysis of the correlation between evaluation criteria and overall satisfaction. Tourism & Management Studies, 10(1), 59–65. Liu, B. (2012). Sentiment analysis and opinion mining. Synthesis Lectures on Human Language Technologies, 5(1), 1–167. Liu, Z., & Park, S. (2015). What makes a useful online review? Implication for travel product websites. Tourism Management, 47, 140–151. Luca, M. (2016). Reviews, reputation, and revenue: The case of Yelp.com (March 15, 2016). Harvard Business School NOM unit working paper, (12–016). Accessed August 8, 2020. Luo, Y., & Xu, X. (2019). Predicting the helpfulness of online restaurant reviews using different machine learning algorithms: A case study of Yelp. Sustainability, 11(19), 5254. Miah, S. J., Vu, H. Q., Gammack, J., & McGrath, M. (2017). A big data analytics method for tourist behaviour analysis. Information & Management, 54(6), 771–785. Miles, M. B., Huberman, A. M., & Saldana, J. (2014). Qualitative data analysis: A methods sourcebook. SAGE.

8

User Comments as a Resource to Rank with Multiple Criteria: The Case. . .

169

Mimno, D., Wallach, H., Talley, E., Leenders, M., & McCallum, A. (2011, July). Optimizing semantic coherence in topic models. In Proceedings of the 2011 conference on empirical methods in natural language processing (pp. 262–272). Nakayama, M., & Wan, Y. (2018). Is culture of origin associated with more expressions? An analysis of Yelp reviews on Japanese restaurants. Tourism Management, 66, 329–338. Newman, D., Noh, Y., Talley, E., Karimi, S., & Baldwin, T. (2010). Evaluating topic models for digital libraries. In Proceedings of the 10th annual joint conference on digital libraries (pp. 215–224). O'Connor, P. (2010). Managing a hotel’s image on TripAdvisor. Journal of Hospitality Marketing & Management, 19(7), 754–772. Ong, B. S. (2012). The perceived influence of user reviews in the hospitality industry. Journal of Hospitality Marketing & Management, 21(5), 463–485. Orlikowski, W. J., & Scott, S. V. (2014). What happens when evaluation goes online? Exploring apparatuses of valuation in the travel sector. Organization Science, 25(3), 868–891. Pang, B., Lee, L., & Vaithyanathan, S. (2002). Thumbs up? Sentiment classification using machine learning techniques. arXiv preprint cs/0205070. Papadimitriou, C. H., Raghavan, P., Tamaki, H., & Vempala, S. (2000). Latent semantic indexing: A probabilistic analysis. Journal of Computer and System Sciences, 61(2), 217–235. Pardalos, P. M., Siskos, Y., & Zopounidis, C. (1995). Advances in multicriteria analysis. Kluwer Academic. Parikh, A., Behnke, C., Vorvoreanu, M., Almanza, B., & Nelson, D. (2014). Motives for reading and articulating user-generated restaurant reviews on Yelp.com. Journal of Hospitality and Tourism Technology. Pohekar, S. D., & Ramachandran, M. (2004). Application of multi-criteria decision making to sustainable energy planning – A review. Renewable and Sustainable Energy Reviews, 8(4), 365–381. Rigney, D. (2010). The Matthew effect: How advantage begets further advantage. Columbia University Press. Rindova, V. P., Martins, L. L., Srinivas, S. B., & Chandler, D. (2018). The good, the bad, and the ugly of organizational rankings: A multidisciplinary review of the literature and directions for future research. Journal of Management, 44(6), 2175–2208. Roy, B. (1968). Classement et choix en présence de points de vue multiples. Revue française d'informatique et de recherche opérationnelle, 2(8), 57–75. Schuckert, M., Liu, X., & Law, R. (2015). A segmentation of online reviews by language groups: How English and non-English speakers rate hotels differently. International Journal of Hospitality Management, 48, 143–149. Sevkli, M. (2010). An application of the fuzzy ELECTRE method for supplier selection. International Journal of Production Research, 48(12), 3393–3405. Shearer, C. (2000). The CRISP-DM model: The new blueprint for data mining. Journal of Data Warehousing, 5(4), 13–22. Siskos, Y., & Spyridakos, A. (1999). Intelligent multicriteria decision support: Overview and perspectives. European Journal of Operational Research, 113(2), 236–246. Steuer, R. E. (1986). Multiple criteria optimization: Theory, computation and applications. Wiley. Taecharungroj, V., & Mathayomchan, B. (2019). Analysing TripAdvisor reviews of tourist attractions in Phuket, Thailand. Tourism Management, 75, 550–568. Titov, I., & McDonald, R. (2008, April). Modeling online reviews with multi-grain topic models. In Proceedings of the 17th international conference on world wide web (pp. 111–120). Tripadvisor. (2020, August 15). Everything you need to know about the Tripadvisor popularity ranking [web page]. https://www.tripadvisor.com/TripAdvisorInsights/w765 Urquhart, C. (2012). Grounded theory for qualitative research: A practical guide. Sage. Wallach, H. M., Mimno, D. M., & McCallum, A. (2009). Rethinking LDA: Why priors matter. In Advances in neural information processing systems (pp. 1973–1981).

170

D. Novas et al.

Wang, X., & Grimson, E. (2008). Spatial latent dirichlet allocation. In Advances in neural information processing systems (pp. 1577–1584). Wiebe, J., Bruce, R., & O’Hara, T. P. (1999, June). Development and use of a gold-standard data set for subjectivity classifications. In Proceedings of the 37th annual meeting of the association for computational linguistics (pp. 246–253). Yan, X., Wang, J., & Chau, M. (2015). Customer revisit intention to restaurants: Evidence from online reviews. Information Systems Frontiers, 17(3), 645–657. Zadech, L. (1965). Fuzzy sets. Information and Control, 8(3), 338–353.

Correction to: Multicriteria Disaggregation- Aggregation Approach for the Evaluation of Warm Water Lakes Dimitrios E. Alexakis, Isaak Vryzidis, and Athanasios Spyridakos

Correction to: Chapter 5 in: A. Spyridakos (ed.), Multicriteria Decision Aid and Resource Management, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-34892-1_5 The book was inadvertently published with incorrect spelling of the author name in chapter 5. The name has been corrected in the chapter.

The updated original version for this chapter can be found at https://doi.org/10.1007/978-3-031-34892-1_5 © Springer Nature Switzerland AG 2023 A. Spyridakos (ed.), Multicriteria Decision Aid and Resource Management, Multiple Criteria Decision Making, https://doi.org/10.1007/978-3-031-34892-1_9

C1