Industrial Symbiosis Recommender Systems

INDUSTRIAL SYMBIOSIS

RECOMMENDER SYSTEMS

Guido van Capelleveen

Theories for the design of recommenders

in complex environments: observations

from industrial symbiosis identification.

Industrial Symbiosis Recommender Systems

PROEFSCHRIFT

ter verkrijging van

de graad van doctor aan de Universiteit Twente, op gezag van de rector magnificus,

prof. dr. T.T.M. Palstra,

volgens besluit van het College voor Promoties, in het openbaar te verdedigen

op woensdag 27 Mei 2020 om 12:45 uur

door

Prof. dr. W.H.M. Zijm (promotor/supervisor) Prof. dr. J. van Hillegersberg (promotor/supervisor) Dr. C. Amrit (co-promotor/co-supervisor)

Dr. D.M. Yazan (co-promotor/co-supervisor)

Woensdag 27 Mei 2020 Wednesday 27th_{May, 2020}

Funding sources:

Ph.D. dissertation, Univerity of Twente Department of Industrial Engineering and Business Information Systems P.O. Box 217, 7500 AE

Enschede, The Netherlands

This research is funded by European Union’s Horizon 2020 research and innovation program under grant agreement No. 680843

SIKS Dissertation Series 2020

The research reported in this thesis has been carried out under the auspices of SIKS, the Dutch Research School for Information and Knowledge Systems.

Cover design by Herlambang Rahmadhani

Printed by Ipskamp Printing, Enschede, The Netherlands Type set with LA_TEX.

ISBN 978-90-365-5007-9

DOI 10.3990/1.9789036550079

Copyright © 2020 Guido van Capelleveen, 2020, Enschede, The Netherlands

All rights reserved. No part of this this dissertation may be reproduced, stored in a retrieval system or transmitted in any form or by any means without permission of the author.

Graduation committee:

Chairman/secretary: Prof. dr. T.A.J. (Theo) Toonen

University of Twente, The Netherlands

First Promotor: Prof. dr. W.H.M. (Henk) Zijm

University of Twente, The Netherlands

Second Promotor: Prof. dr. J. (Jos) van Hillegersberg

University of Twente, The Netherlands

Co-promotor: Dr. C. (Chintan) Amrit

University of Amsterdam, The Netherlands

Co-promotor: Dr. D.M. (Devrim) Yazan

University of Twente, The Netherlands

Committee members: Prof. dr. M.E. (Maria) Iacob

University of Twente, The Netherlands Prof. dr. E. (Evangelos) Kanoulas

University of Amsterdam, The Netherlands Prof. dr. J.N. (Joost) Kok

University of Twente, The Netherlands Dr. M. (Miquel) S`anchez-Marr`e

Universitat Polit`ecnica de Catalunya-BarcelonaTech, Spain Prof. dr. R.J. (Roel) Wieringa

Acknowledgements

It all started a few years ago in Brazil. During my visit I spoke with one of my friends, Vinicius, about his career in science which became an inspiration to explore such possibility. Even though I worked for some time in business, I decided to come back to university life and fully commit to a PhD. Fortunately, this journey turned out to be an enriching experience and also resulted in a thesis. However, this dissertation would not have been completed without the support of many others. In the following paragraphs I would like to express my gratitude towards all of you, supervisors, colleagues, friends and family that have supported me in their own ways.

First I would like to thank my promoters Henk and Jos, and my supervisors Devrim and Chintan, for their scientific review contributions to this thesis, their efforts on personal development, and their confidence in me to excel in my position at the UT. I am particular thankful to my first promotor Henk for his critical review and detailed feedback on my work. It is worthy to mention that even after a long career you were still driven to dive into a multi-disciplinary project which required you to learn about concepts outside your core research area. I would like to thank my second promotor Jos, not in the first place for introducing me to the PhD position around 4,5 years ago, but also for always seeing new opportunities during the PhD to develop other academic skills in addition to thesis writing. A warm gratitude goes out to Devrim. It has been a pleasure to work with you and have you as my supervisor. The passion you have for your work is tremendous and you were always there to manage the project, classes or my progression. We have travelled together around Europe to visit industries, project meetings and experienced other cultures. I particularly enjoyed our moments we went for dinner or drinks during which we discussed life. Furthermore, I am very grateful to Chintan. I appreciate you for your lively enthusiasm and laughs every time you enter the room. I am grateful for your ability to structurally discuss the theoretical contents and let me think about the ’story selling’ of my work. Your ambition challenges me to push forward where I should be. My gratitude also goes to the opponents in my graduation committee, Evangelos, Joost, Maria, Miquel, and Roel, for reviewing my thesis and providing me the honorable opportunity to defend this work.

A profound thanks also goes to my project partners Vahid and Luca. The intensive collaborations, discussions and of course the fun we had in the SHAREBOX project will not be forgotten. Working with both of you has been fruitful, educational and moreover a pleasure. Furthermore, I would like to express my gratitude to the other authors of papers published as part of this theses, Jesse and Asad, thank you for your welcome additions in Chapter 4 and 5. Besides the thesis writing, it was also a privilege to collaborate with a number of people on slightly different academic subjects. Olcay, thank you for the opportunity to collaboratively work on social network theory and bio-mimicry for industrial symbiosis. Johanna, Andreas, and Daniel thanks for our discussions on improving life-cycle data management. Tiffany thank you for our work on detecting Islamophobia using machine learning and trend analysis.

Much of the time spend at the office has been a pleasure because of my IEBIS colleagues. I enjoyed the discussions with all I shared our flex-office, among who are Abhishta, Andrej, Atil, Iqbal, Leon, Lucas, Luka, Miha, Olcay, Robin, Robert, and Vahid. In addition, I also highly appreciate the interesting discussions during lunch with Andrej, Asad, Berend, Berend, Fabian, Gr´eeanne, Martijn, Miha, Mike, Olivia, Reinoud, Rogier, Sajjad Sebastian, Sjoerd, Wenyi, and Xavier. Special thanks to Abhishta, Letizia, Wouter, Nils, Laura, Arturo, and other members of the group that occasionally joined for the night outs in Rico, Rocks or Molly, and the many other outbreak activities we had together in Enschede. Thank you Nils for our work-out sessions in the Cube; being able to exercise helps to keep up the work ethic. Abhishta, thanks of course for your delightful cooking everytime. Finally, also a big thank you to Elke, Hilde, and Gea for all the support provided in arranging trips, meetings and other organizational tasks.

My appreciation furthermore goes out to the friendships made all around the globe during my time as a doctoral candidate (and of course also before that). I always enjoyed and will enjoy to have your company, even when situations are limiting our physical ability to do so. No matter whether the activity involved riding the slopes with you in Austria, France or Switzerland; kayaking so many rivers in the Alps; steak travelling to Estonia; BBQ’ing in Utrecht; forest overnights in Siegen; visiting you in Warsaw, Dublin, or elsewhere; motorbiking through the Eiffel; bouldering, climbing, or zip-lining; online alignment sessions; mountain biking; camping; or any other trip we made, these fine memories will last forever. A special thanks to all of you at Euros Kano and the kayak community as a whole. You made me feel so at home that I could regain the energy to work on my thesis. I have spend much of my time being on the road or on the water with you, and couldn’t have finished a thesis without having such a close group of friends that

iii

share my passion. I hope we are able to continue our journey and find you back on the water soon. Thank you all for having me!

Last but not least, pursuing a PhD would not be possible without the support of my family, Eric, Corrie, Renske and Tobias. Each of you provided me with inspiration, knowledge, discussions, mental support or sometimes the necessary distractions. Utmost, I would like to express my gratitude to Eva, for her devotion, care and patience in the finalizing months of writing this manuscript. Your smile can produce miracles.

For a long time, humanity has lived upon the paradigm that the amounts of natural resources are unlimited and that the environment has ample regenerative capacity. However, the notion to shift towards sustainability has resulted in a worldwide adoption of policies addressing resource efficiency and preservation of natural resources. One of the key environmental and economic sustainable operations that is currently promoted and enacted in the European Union policy is Industrial Symbiosis. In industrial symbiosis, firms aim to reduce the total material and energy footprint by circulating traditional secondary production process outputs of firms to become part of an input for the production process of other firms.

This thesis directs attention to the design considerations for recommender systems in the highly dynamic domain of industrial symbiosis. Recommender systems are a promising technology that may facilitate in multiple facets of the industrial symbiosis creation as they reduce the complexity of decision making. This typical strength of recommender systems has been responsible for improved sales and a higher return of investments. That provides the prospect for industrial symbiosis recommenders to increase the number of synergistic transactions that reduce the total environmental impact of the process industry in particular [1].

This exploratory recommender system work is a first attempt in the field of industrial symbiosis. The work highlights three designs that justify the application for recommender systems to improve decision making in information systems facilitating the industrial symbiosis identification process. Our first case [2] shows a sector-based recommender system design that supports the identification of industrial symbiosis opportunities to industries that are, in general, unaware of the potential of industrial symbiosis. This recommender’s aim is to attract a critical mass of industries to actively participate in a new exchange network. The initial burden for the recommender is that industries are not yet convinced to provide detailed data and are not committed to invest substantial resources for investigating potential business. Once such a network is established, we find the application of two other recommenders. The second case [3] shows a recommender system design that supports the identification of the correct European Waste Catalogue (EWC) code when listing a waste item in an industrial symbiosis market platform used by the network in which the industry participates. This recommender is challenged

v

by the interpretation of natural language and the EWC classification system used in industrial symbiosis applications throughout Europe. Finally, the third case [4] shows a recommender system design which supports industries in the identification of waste items of interest. This recommender is faced with the data representation and confidentiality issues that arise when enlisting wastes in a marketplace.

The numerous burgeoning concepts in recommender system development, the variety of choices, and the numerous alternative system designs, may be overwhelming for practitioners to consider. Hence, recommender system designers would benefit from a structured ontological approach in coordinating design

decisions. Guided by the observations from the three industrial symbiosis

recommenders, and supported by the existing academic knowledge base, we formalize a generalized framework to approach the design of recommender systems in complex domains. This framework [5] is proposed as a canvas management tool to facilitate recommender system design. It provides a novel requirement specification perspective on recommender systems which enables practitioners to create a high-level structured overview of recommender system designs while externalizing the relationships between interrelated concepts.

Voor een lange tijd heeft onze samenleving geleefd volgens het paradigma dat natuurlijke grondstoffen ongelimiteerd beschikbaar zijn en de aarde over voldoende regeneratieve capaciteit beschikt om blijvend aan onze grondstofbehoefte te voldoen. Echter, in de loop der jaren is het besef toegenomen dat een duurzame en waar mogelijk circulaire economie een beter toekomstperspectief biedt. Dit heeft zich vertaald in de vele beleidsregels die op alle bestuurslagen zichtbaar zijn geworden, waaronder die in de Europese Unie. Een aantal van deze beleidsregels richt zich op het fenomeen industri¨ele symbiose. Het doel van industri¨ele symbiose is om het totale grondstof- en energieverbruik te reduceren door de traditionele secundaire uitvoer van het productieproces (d.w.z. de reststoffen en restenergie) van bedrijven te (her)gebruiken als een invoer van productieprocessen bij andere bedrijven.

Dit proefschrift richt zich op de aanpak en ontwerpoverwegingen ten behoeve van aanbevelingssystemen in het dynamische domein van de industri¨ele

symbiose. Aanbevelingsalgorithmen vormen een veelbelovende technologie

die de totstandbrenging van industri¨ele symbiose in vele facetten zou kunnen ondersteunen doordat het de complexiteit van beslissingen kan reduceren. Het is deze typerende kracht van aanbevelingssystemen die zich in vele andere domeinen al heeft bewezen met als resultaat een verhoging van de omzet en een verhoogd rendement op de investering. Dit doet vermoeden dat aanbevelingssystemen ook een positief effect zouden kunnen bewerkstelligen in het verbinden van industrie¨en door middel van synergie¨en in het (her)gebruik van grondstoffen en energie. Dit kan resulteren in een gewenste reductie van de totale impact van bedrijven op het milieu, in het bijzonder van bedrijven in de procesindustrie.

Deze exploratieve studie gericht op het ontwerp van aanbevelingssystemen is een eerste poging in het werkveld van de industri¨ele symbiose. In het proefschrift worden drie ontwerpen toegelicht die de toepassing van aanbevelingssyteem technologie in informatiesystemen rechtvaardigen door het vereenvoudigen van

beslissingen in het identificatieproces van industri¨ele symbiose. De eerste

casus toont een sectorgebaseerd aanbevelingssysteem ontwerp dat faciliteert in de vroegtijdige identificatiefase van industri¨ele symbiose gekenmerkt door een industrie die zich nochtans onbewust is van het potentieel dat industri¨ele

vii

symbiose kan bieden. Het doel van dit aanbevelingssysteem is dan ook om

een kritieke massa van industrie¨en aan te trekken die actief willen participeren in een netwerk waarin restgrondstoffen en energie zouden kunnen worden uitgewisseld. De centrale uitdaging voor het aanbevelingssysteem is dat industrie¨en in eerste instantie nog niet erg zijn genegen om gedetailleerde gegevens te verstrekken en bovendien nog niet vastbesloten zijn substantieel te investeren in onderzoek naar deze nieuwe vorm van potenti¨ele bedrijvigheid. Echter, zodra een dergelijk netwerk is gerealiseerd, zien we mogelijkheden voor twee andere aanbevelingssystemen. De tweede casus toont een aanbevelingssysteem ontwerp dat de identificatie van een correcte Europese Afval Catalogus (EWC) code ondersteunt. Deze aanbeveling vind plaats in de processtap waarin een afvalproduct in een marktplatform voor industri¨ele symbiose wordt geregistreerd. De centrale uitdaging voor het aanbevelingssysteem is de interpretatie van natuurlijk taal en de karakteristieken van het EWC-classificatiesysteem dat de facto standaard is voor industri¨ele symbiose applicaties in Europa. Ten slotte toont de derde casus een aanbevelingssysteem ontwerp dat industrie¨en ondersteunt bij de identificatie van afvalproducten in de regio die passen bij de interesses van de desbetreffende

procesindustrie. De centrale uitdaging voor dit aanbevelingssysteem is de

representatie van gegevens en vertrouwelijkheid van informatie die gepaard gaan met de registratie van afvalproducten in een markt voor industri¨ele symbiose.

De talloze snelgroeiende concepten in het ontwerp van aanbevelingssystemen, de verscheidenheid aan keuzes, en de vele alternatieve systeemontwerpen kunnen overweldigend zijn voor de systeem ontwerpers om te overwegen. Vandaar dat aanbevelingssysteem ontwerpers zouden kunnen profiteren van een gestructureerde ontologische benadering bij het co¨ordineren van al deze ontwerpbeslissingen. Gestuurd door de observaties van de drie aanbevelingssystemen voor industri¨ele symbiose, en ondersteund door de bestaande academische kennisbasis, biedt dit proefschrift een geformaliseerd algemeen kader om het ontwerp van aanbevelingssystemen in complexe domeinen te benaderen. Dit raamwerk wordt voorgesteld als een canvas beheersinstrument om aanbevelingssysteem ontwerp te vereenvoudigen. Het biedt een nieuw perspectief voor het opstellen van een programma van eisen aan aanbevelingssystemen die ontwerpers in staat stelt om een gestructureerd overzicht op hoog niveau van aanbevelingssysteem ontwerp te cre¨eren alsmede om de relaties tussen onderling verbonden concepten te analyseren.

Contents

1.1.1 Emergence of industrial symbiosis . . . 2

1.1.2 Organization of industrial symbiosis . . . 3

1.1.3 Drivers and barriers in industrial symbiosis . . . 4

1.2 Recommender systems for industrial symbiosis . . . 4

1.2.1 Recommender supported identification of industrial symbiosis . . . 5

1.2.2 The need for a structured design approach of industrial symbiosis recommender systems . . . 6

1.3 Organization of the thesis . . . 8

2 A review of research gaps 15 2.1 Introduction . . . 15

2.2 Methodology . . . 16

2.3 Industrial symbiosis tools . . . 18

2.3.1 Open online waste markets . . . 24

2.3.2 Facilitated synergy identification systems . . . 25

2.3.3 Industry sector synergy identification . . . 27

2.3.4 Social network synergy identification . . . 28

2.3.5 Industrial symbiosis knowledge repositories . . . 28

2.3.6 Industrial symbiosis region identification . . . 29

2.4 IT challenges for industrial symbiosis tools . . . 31

2.4.1 Building decision support . . . 31

2.4.2 Providing recommendation . . . 31

3 The research approach 37

3.1 Project objectives and research aims . . . 37

3.2 Design goal . . . 38

3.3 Knowledge goal and questions . . . 39

3.4 Research methods . . . 40

4 A recommender system for the identification of industrial symbiosis ideas: exploration through interactive visualization 43 4.1 Introduction . . . 43

4.2 Related works . . . 45

4.3 Methodology . . . 47

4.3.1 Problem identification and motivation . . . 48

4.3.2 Objectives for a solution . . . 48

4.3.3 Artifact design . . . 49 4.3.4 Artifact demonstration . . . 49 4.3.5 Artifact evaluation . . . 50 4.3.6 Communication . . . 50 4.4 Artifact design . . . 51 4.4.1 Data collection . . . 51 4.4.2 Item prediction . . . 54 4.4.3 Hierarchical exploration . . . 57 4.5 Artifact demonstration . . . 61 4.6 Artifact evaluation . . . 65 4.7 Discussion . . . 66 4.8 Conclusion . . . 69

5 A tag recommender system for the circular economy: exploring the perils of the European Waste Catalogue 73 5.1 Introduction . . . 73

5.2 Background . . . 75

5.2.1 Enhancing data for tag recommendation . . . 75

5.2.2 Measuring Short Text Semantic Similarity . . . 76

5.3 Methodology . . . 77

5.3.1 Problem definition . . . 77

5.3.2 Data characteristics of the ’waste’ domain . . . 78

5.3.3 Experiment setup . . . 81

5.3.4 Proposed methods . . . 83

5.4 Metrics and performance evaluation . . . 87

5.4.1 Word2vec model and hyperparameters . . . 87

CONTENTS xi

5.4.3 Parameter setting of Recommenders . . . 93

5.4.4 Comparison of methods . . . 97

5.5 Discussion . . . 100

5.5.1 Causes for incorrect EWC tag recommendation . . . 100

5.5.2 EWC Tag Ontologies . . . 104

5.5.3 Limitations . . . 105

5.6 Conclusion . . . 105

6 A recommender system to facilitate industrial symbiosis markets: how knowledge influences recommender design 109 6.1 Introduction . . . 109

6.2 Methodology . . . 111

6.2.1 Step 1: Data collection . . . 111

6.2.2 Step 2: Model design . . . 112

6.2.3 Step 3: Instantiation of the model . . . 112

6.2.4 Step 4: Design evaluation . . . 113

6.3 Model design: recommendation in an environmental data landscape 114 6.3.1 The model . . . 114

6.3.2 The model applied to the case of Industrial Symbiosis . . 115

6.3.3 Pre-processing IS data . . . 119

6.3.4 Design of an explicit knowledge based IS recommendation 121 6.3.5 Design of an implicit knowledge based IS recommendation 124 6.4 Design evaluation . . . 125

6.4.1 Data preparation . . . 125

6.4.2 Results of recommender performance . . . 126

6.5 Discussion . . . 129

6.5.1 The role of explicit and implicit knowledge . . . 129

6.5.2 Data challenges to explicit knowledge based industrial symbiosis recommenders . . . 131

6.5.3 Internal and external validity . . . 133

6.6 Conclusion . . . 134

7 The recommender canvas 139 7.1 Introduction . . . 139

7.2 Methodology . . . 141

7.3 The recommender system design model . . . 143

7.3.1 Goals . . . 144

7.3.2 Domain characteristics . . . 147

7.3.3 Functional design considerations . . . 151

7.3.5 Interface design . . . 161

7.3.6 Evaluation and optimization . . . 166

7.4 An example case of the recommender canvas: Sharebox . . . 168

7.5 Discussion . . . 171

7.5.1 Use of Recommender System Models . . . 171

7.5.2 Internal & External validity . . . 172

7.6 Conclusion . . . 173

8 Conclusions 175 8.1 Revisiting the design goal and knowledge questions . . . 175

8.2 Notes for discussion . . . 179

8.2.1 Controversies around industrial symbiosis affecting the recommendation . . . 179

8.2.2 Impaired user acceptance of decision support . . . 181

8.2.3 Impact of recommender systems on the diversity of choice 182 8.3 Supplementary research . . . 184

8.3.1 Improving the collection, storage and distribution of life cycle inventory data for industrial symbiosis . . . 184

8.3.2 Agent-based simulation of industrial symbiosis . . . 184

8.3.3 Eco-industrial parks development . . . 185

8.3.4 Energy-based industrial symbiosis . . . 186

8.4 Directions for future research . . . 186

Appendices 187 A Case I Questionnaire . . . 189

B Case III Algorithm: Stem frequency vectorization . . . 190

C Case III Algorithm: A simple multi-dimensional hierarchical agglomerative clustering algorithm . . . 191

Bibliography 192

About the author 240

List of publications 241

CHAPTER 1

Introduction

This chapter introduces the background and motivation for the research presented in this thesis. We discuss the concepts of industrial symbiosis and recommender systems and explain why the latter may play a key role in exploring industrial symbiosis opportunities. The chapter ends with an overview of the organisation of this thesis and lists the key contributions per chapter.

1.1

Industrial symbiosis

The reduction of waste emissions and primary resource use in resource-intensive industries is suggested as one of the critical pathways to accelerate sustainable development [6]. The field of industrial ecology, which studies the physical flows of substances and transformation processes in organizations, offers a perspective on how to reduce waste in industrial systems by transforming linear system thinking into circular system thinking. One of the most commonly accepted early definitions of this industrial ecosystem thinking was proposed by Frosch and Gallopoulos as: ”the transformation of the traditional model of industrial activity, in which individual manufacturing takes in raw materials and generates products to be sold plus waste to be disposed of, into a more integrated system, in which the consumption of energy and materials is optimized and the effluents of one process

serve as the raw material for another process” [7]. The result of conceptual

modeling an industrial ecosystem with its circular links from and to the natural ecosystems contributed to the evolutionary thinking on interrelated symbiotic links among groups of firms within the ecosystem, which is now known as industrial symbiosis.

Industrial Symbiosis (IS) entails the identification and utilization of an

organization’s traditional secondary production process output that is generally considered as waste. Such waste can be used to substitute (part of) a primary resource (possibly after some pre-processing) in the production process of another

organization, usually representing a different industrial sector [8]. Such a

transaction is often referred to as a synergy between the industries. The industrial symbiosis methodology is gaining popularity as a means to improve sustainable

production because of its ability to provide economic gains while yielding ecological and social benefits [9]. Both researchers and practitioners recognize this impact of industrial symbiosis as a useful business opportunity and tool for eco-innovation. This practitioners perspective led to describe industrial symbiosis as ”engaging diverse organizations in a network to foster eco-innovation and long-term culture change, creating and sharing knowledge through the network yielding mutually profitable transactions for novel sourcing of required inputs, value-added destinations for non-product outputs, and improved business and technical processes” ([10], page 28). Because of the synergy between the vast amount of positive environmental effects and the economic benefits for industries, it is not surprising that the European Union highlighted industrial symbiosis as a methodology to stimulate industries to become more sustainable [11].

1.1.1

Emergence of industrial symbiosis

Many applications of industrial symbiosis arose since the case of the highly cited self-organizing Danish industrial town of Kalundborg, one of the early notorious industrial symbiosis cases with an apparent success reported in the academic literature [12], [13]. The industrial area of the city, which has been developing since the 1960s, became wider known in 1989 when Christensen provided insight into the ecosystem of Kalundborg and coined the term of industrial symbiosis [14]. However, sceptics point out the controversy of eco-industrial parks. In their view industrial symbiosis is rather a re-discovery of known economical principles [15]–[17]. Other scholars argue that the key aspect of industrial symbiosis is not the reference it makes to the natural existence of symbiotic relationships but that the design of the industrial ecosystem is built from scratch, i.e, the specific industries and their location are carefully chosen when composing an industrial park configuration [18]. Nevertheless, many lessons have been learned from the Kalundborg area and numerous efforts are made trying to replicate the formula [19]. It is argued that these cases play a central role in the discussion to transfer theoretical principles of industrial ecology to practice and form examples for today’s practices [20]. Shortly after Kalundborg, many similar projects could be observed in, among other places, Jyv¨askyl¨a in Finland [21], Upper Styria in Austria [22], the Rotterdam harbour [23] and the western Australia mineral mining region [24], to be rapidly spread over the world [25]–[34]. As of today, hundreds of applications are known in the literature [35]–[37] under the term of industrial symbiosis, and possibly even more under different process terminology [38]. In some nations the emergence of industrial symbiosis even led to stimulation programs, including the National Industrial Symbiosis Program (NISP) in the United Kingdom and the Landskrona program in Sweden [39], [40]. Authorities

1.1. INDUSTRIAL SYMBIOSIS 3

are also taking a key role in the development of industrial symbiosis by developing policy interventions and facilitating sustainable development [41], such as the European policy [11], [42]. Policy instruments allow authorities to discourage waste disposal and provide incentives to make industrial symbiosis economical attractive and sustainable [27], [43]–[45]. In some projects a government may even take the role of infrastructure provider [46], [47].

1.1.2

Organization of industrial symbiosis

The emergence of industrial symbiosis is intensively studied by industrial ecology researchers. They propose various models which attempt to explain the emergence of industrial symbiosis by characterizing the development using typical dynamics, such as the environment, the initial actors, motivations and the outcomes of industrial symbiosis actions [48]. The most prevalent models are the planned, self-organized and facilitated model [9], [49]. These models not only serve to understand the complexity of industrial symbiosis, but also contribute in the search for common driving forces and evolutionary trajectories.

The planned model is a continuous effort from coordinating bodies, typically municipalities or regional governments, to relocate industries that can exchange wastes in an industrial park region [27]. The self-organized model results from collaboration without top-down planning and is mainly driven by economic or strategic business motivations that lead to an increased number of resource and waste transactions over time [50]. The facilitated model utilizes intermediaries that fulfil the role of connecting industries through synergy identification, strengthening the trust between firms and consulting the firms with their expertise to configure

sustainable synergies [49]. These intermediaries can be part of an executive

government, such as in the subsidized case of the Landskrona Industrial Symbiosis Program (LISP) in Sweden [39], [40], or are business driven facilitators such as in the National Industrial Symbiosis Program (NISP) in the United Kingdom [49]. These pathways not only characterize the different types of emergence and explain how the process of industrial symbiosis unfolds, but also help to deduce the critical catalyzers to initiate new symbiotic actions [48].

In the decision making of procurement of raw materials and the acquirement of industrial capacity, products or services, industrial symbiosis can be considered as an alternative approach for traditional suppliers. However, industrial symbiosis typically requires a prior investigation related to, e.g., the alternative material consideration, the economic viability, the legal issues, the logistics, potential material conversion requirements, risks mitigation, and an evaluation of potential

buyer-supplier bargaining power change [51]. From a process point of view

1. Identification of the synergy and the potential eligible industries (e.g., searching for alternative materials, identifying industrial partners).

2. Evaluation of the feasibility of the synergy (e.g., identifying economic and environmental benefits, testing the technology readiness, assessing the sustainability of the collaboration, allocating the cost, risk analysis, checking legal issues, etc.).

3. Implementation and coordination of the synergy (e.g., contracting, arranging transport, deploying required conversion technology, etc.). 4. Monitoring the synergy (e.g., collecting synergy data, re-assessing the

sustainability of a synergy or configuration, etc.).

1.1.3

Drivers and barriers in industrial symbiosis

With the experience of both success and failure, numerous studies examine the

drivers and barriers of industrial symbiosis (e.g., [38], [55], [56]). The key

drivers for industrial symbiosis include mutual economic benefits, low proximity or co-location of opportunities, well-formed government regulations, a strong anchor industry, a diversity of actors in a network, and a common strategic believe. The intermediaries, coordinators and pioneers may be viewed as important drivers as these develop trust, openness, knowledge, embeddedness, a social and cultural network along with effective communication and governance monitoring. On the other hand, the key barriers for industrial symbiosis include the power, status and asymmetries of organizations, overly broad diversity among partners, the exit of partners, excessive costs and risk, restrictive environmental regulations and the lack of trust among firms. Most industrial symbiosis transactions however are driven by economic principles, e.g., the mutual profit, revenue enhancement or business expansion for firms. Hence, the assessment of the sustainability, the environmental or the social impacts of a synergy usually is only of secondary interest for firms [57].

1.2

Recommender systems for industrial symbiosis

The consideration to support managerial work, in particular to recognize, structure, assess and solve (strategic) management problems, is today inherent to the power of data, computation and knowledge-driven reasoning. Systems that support work involving managerial decision making are addressed in the field of decision support systems [58], [59], with a sub field of recommender systems [60] which is the

1.2. RECOMMENDER SYSTEMS FOR INDUSTRIAL SYMBIOSIS 5

computer-based information system that can process, manipulate and present data with the help of tools that aid in the decision making process in unstructured and semi-structured situations when managerial judgment is required [61]. A principle for providing decision support is that it presupposes a detailed understanding of the manager’s or organizational decision problem [62]. To develop this knowledge one should identify the work processes that need management and gather data from the potential end-users in order to formulate the decisions that are to be made such that the task support can be identified.

The typical purpose of a Recommender System (RS) is to actively suggest items of interest to users. Thereby, recommender systems facilitate the exploration of available options while also filtering a set of items in an attempt to decrease the users’ information load [63]. Recommender systems are found today in many applications [64]. For example, recommender systems are present in search engines to suggest news [65], in social networks designed to recommend friends [66], while other systems support passengers to locate vacant taxis [67], suggest travel offers to tourists [68], support coaches to identify the athletes who begin losing motivation [69], and match process industries based on waste re-use potential [4]. Its popularity both in research and practice is evident in the various literature surveys that provide a comprehensive summary of the recommender system field [70]–[75].

1.2.1

Recommender supported identification of industrial

symbiosis

The identification phase of industrial symbiosis, where organizations identify the most promising areas and synergies to be turned into a business case, are still experienced as a resource intensive process. For example, ’company A states it wants to dispose large amounts of a particular waste sludge produced on a daily basis (specific details typically intentionally omitted). While the industry currently is disposing the waste to landfill, it is interested to learn which types of symbiosis the waste could be converted into, and which are the (required) local partners (industry types) to construct the symbiotic transfer with’. Among others, industrial symbiosis opportunities are typically identified during facilitated industry workshops [49], [76], [77], by using specific industrial symbiosis identification systems [1], [52], [78]–[80], and through exploration of items in industrial symbiosis marketplaces [81]. Such systems can enhance symbiotic transactions in the network while substantially reducing the time investment required to

investigate the symbiotic potential. The capability of supporting the learning

and decision-making processes of environmental problems recurs as an interesting opportunity to assess [82]. Therefore, decision support tools [52] and in particular

recommender support are suggested as promising techniques to stimulate and facilitate the identification and assessment of new exchanges. A detailed overview of the different ICT tools that facilitate industrial symbiosis can be found in the background Chapter 2 which indicates there is still a clear lack of effective decision support in these systems, and in particular illustrates the gap for researching industrial symbiosis recommenders.

Recommenders are able to support users in identifying item opportunities and to pro-actively engage system use, resulting in both increased sales and a more active community [83]–[85]. A recommender system can contribute to the reduction of search costs associated with identifying relevant industrial symbiosis ideas for a firm, support the firm in classifying waste items with EWC code labels, or suggest the waste items that match the firm’s waste preference profile. There are however various challenges when building systems that can provide decision support or recommend industrial symbiosis opportunities [52]. For example, a precondition for the success of a waste marketplace is that a critical mass of industries must first be engaged to actively participate in the network prior to operating any industrial symbiosis recommender or decision support tool that can reason on the basis of specific transactions to be made [52]. Moreover, industrial symbiosis identification systems struggle with analyzing data because of the high level of implicit knowledge in industrial symbiosis, which is a burden to the development of techniques that help to identify industrial symbiosis exchanges [52]. For example, process data from manufacturing industries that disclose inputs, outputs, and wastes, can be both used for the identification of potential synergies as well as provide input for recommender systems. However, the extent and level of detail with which such data is shared may be hindered by a lack of trust among organizations because process data might reveal competitive information that organizations want to keep (partly) confidential [49]. Moreover, organizations have to justify time and resource investments to explore potential ideas for which the expected benefits are not clearly predicted or even known. Therefore, the provision of detailed process data in a wider context may increase identification of industrial

symbiosis opportunities, but this is challenging at the initial stage. This data

challenging situation describes the framework in which we design recommender systems that assist the identification of industrial symbiosis which is characteristic to all the cases on recommender design presented in this thesis.

1.2.2

The need for a structured design approach of industrial

symbiosis recommender systems

The design of a recommender system is a challenging proposition due to numerous factors. A few of these include: data quality [86], the prioritization

1.2. RECOMMENDER SYSTEMS FOR INDUSTRIAL SYMBIOSIS 7

of recommender system goals (e.g., precision, recall, accuracy, and novelty of item suggestions) [70], the increase in size [70], more complex problem domains

[87] and more diverse applications across various domains [88]. In addition,

new concepts such as trust derived from social networks [89], context-awareness enabled through increased mobile phone use [68], and the need for configurable human-recommender interaction [90], have also added design considerations when building recommender systems. As a result of all these contextual features and design features engineers have begun to design more distinctive and advanced (hybrid) filtering techniques.

The challenges around data, a contextual feature, in relation to other design features (e.g., the filtering algorithm design, dimensionality reduction techniques, evaluation metrics, interaction design, item explanation design) are the key focus in the cases on recommender systems for industrial symbiosis identification studied. The cases studied and presented in this thesis are identified based on three problems that operations managers face in their role as a process industry pursuing new industrial symbiosis opportunities which could be supported by a recommender. To begin with, a recommender system could help to identify relevant industrial symbiosis ideas for a firm, addressed as Case I. Next, a recommender system might support a firm in classifying waste items with EWC code labels, addressed as Case II. Finally, a recommender system could suggest the waste items in an established waste marketplace that matches the firm’s waste preference profile, addressed as Case III.

These data challenged recommender designs form the inspiration and provide the initial basis for developing our central theory on recommender design placed in the broader framework of recommender systems. That is, we suggest there exist relationships (i.e., dependencies and (bi-)directional influences) between the various design concepts in recommender design, which make a particular strong or weak recommender system design. These design concepts should be captured during the design of a recommender system. This thesis proposes an ontological model (i.e., the recommender canvas in Chapter 7) to support engineers

in designing recommender systems for more complex domains. Following a

hermeneutic literature review approach [91], focused on the process of developing an understanding, while respecting the creativity of design in a large body of literature relevant to the target problem, we capture and structure all design concepts that can help to analyze these relationships between the features during the design phase of a recommender system in industrial symbiosis.

1.3

Organization of the thesis

This thesis consists of eight chapters. Figure 1.1 shows how the thesis is organized and how the chapters are related. Below we provide a brief overview for each thesis chapter and list the reference(s) to the peer-reviewed publications on which the chapter is based.

Figure 1.1: Organization of the thesis

Chapter 1: Introduction

The introduction chapter clarifies the motivation for the thesis, first by explaining the concepts of industrial symbiosis and recommender systems, followed by an explanation how the latter might be used to explore industrial symbiosis opportunities and help to implement them.

Chapter 2: A review of research gaps

This background chapter presents the literature to illustrate the scientific gap of recommender systems in industrial symbiosis. It analyzes information systems which act as a facilitator of communication and distributor of knowledge and describes (i) the characteristics of these different information systems, (ii) the role of support these systems provide, and (iii) the technologies used to enable such identification. The chapter analyzes the current state of the art in the literature that addresses information systems that facilitate industrial symbiosis identification and studies the systems using these three pillars mentioned above. This chapter contributes by providing a classification framework of information systems that

1.3. ORGANIZATION OF THE THESIS 9

facilitate industrial symbiosis identification and reveals three research directions to progress industrial symbiosis identification tools, being (i) software product and service development, (ii) data integration, and (iii) adoption of intelligent learning with in particular recommender systems. The background chapter is based on the peer-reviewed publication:

• G. van Capelleveen, C. Amrit and D. M. Yazan, ‘A literature survey of information systems facilitating the identification of industrial symbiosis’, in From Science to Society: New Trends in Environmental Informatics, B. Otjacques, P. Hitzelberger, S. Naumann and V. Wohlgemuth, Eds.

Cham: Springer International Publishing, 2018, pp. 155–169, ISBN:

978-3-319-65687-8, [Online]. Available: https : / / doi . org / 10 . 1007 / 978-3-319-65687-8_14

Chapter 3: The research approach

The research approach chapter defines the research goals, formulate the research questions, and describes the research methodologies used in the thesis. While the core question of this thesis relates to recommender design, it finds application in the research domain of industrial symbiosis. Therefore, we specify the overall research question on recommender system design into three sub questions that find application in the identification of industrial symbiosis. In addition, for each of the questions we explain at a high level the selection of methods to investigate the sub problems and show how these contribute to the development of our main theory.

Chapter 4: A recommender system for the identification of industrial

symbiosis ideas: exploration through interactive visualization

This chapter presents the first case investigated which is a sector-based recommender system that supports the identification of industrial symbiosis opportunities to industries that are, in general, unaware of the potential of industrial symbiosis. The chapter focuses on the paradigm of exploring recommendations through interactive visualization as this may enhance users with more control over recommendations and increase transparency of the recommendation process which is a rising need expressed by users of recommender systems. Which interactive visualization method to select is a task for which there still is limited scientific knowledge, especially why a certain method has a strong capacity to support the exploration of recommendations. Therefore, a nascent field of design science oriented research addresses case applications of interactive exploration methods

in recommender systems. We focus on the interactive node-link visualization

use as present in industrial symbiosis. We design, implement and evaluate a recommender system (the IS Identification App) that supports sector-based

identification with the application of interactive node-link visualization. This

case indicates that the interactive hierarchical visualization technique is a viable, fast and effective approach for exploring recommendations and contributes to the practice of interactive node-link visualization for exploring industrial symbiosis recommendations. The chapter further illustrates the effectiveness of this approach, and provides process industries with a promising sector-based recommender approach to identify industrial symbiosis opportunities. This chapter is based on the following submitted manuscript for peer-reviewed publication:

• G. van Capelleveen, J. van Wieren, C. Amrit, D. M. Yazan and H. Zijm, ‘Interactive hierarchical visualization to support exploration in a recommender system for the identification of industrial symbiosis’, Decision Support Systems, 2020 (Under review)

Chapter 5: A tag recommender system for the circular economy: exploring the perils of the European Waste Catalogue

This chapter presents the second case investigated which is a recommender system design that supports the identification of the correct European Waste Catalogue (EWC) code. This chapter focuses on the EWC coding system because the growing number of industries aiming at more sustainable business processes drive the use of the European Waste Catalogue (EWC) in application areas other than originally intended. Such an area is the identification of industrial symbiosis opportunities in which a user-generated item description has to be annotated with exactly one EWC tag from an a priori defined tag ontology consisting of EWC codes. The current study aims to help researchers and practitioners understand the perils of the EWC when building a recommender system based on natural language processing techniques. Direct term matching algorithms generally fail to retrieve the relevant EWC tag because there is often no identical term found in the typically short text generated by users. To increase the relevance of tag recommendation we experiment with semantic enhancement (an EWC thesaurus) and the linguistic contexts of words (learned by Word2vec) in detecting term vector similarity. In contrast to our expectations, the performance improvement is only marginal. Our in-depth analysis shows insights into why the different recommenders were unable to generate a correct annotation. This poses a discussion on the current design of the EWC system for its use in applications facilitating sustainability for process industries. Moreover, it provides directions for practitioners on how to improve EWC algorithms. This chapter is based on the following submitted manuscript for peer-reviewed publication:

1.3. ORGANIZATION OF THE THESIS 11

• G. van Capelleveen, C. Amrit, H. Zijm, D. M. Yazan and A. Abdi, ‘Towards building recommender systems for sustainable industries: The perils of the european waste catalogue’, Journal of Environmental Management, 2020 (Under review)

Chapter 6: A recommender system to facilitate industrial symbiosis markets: how knowledge influences recommender design

This chapter presents the third case investigated which is a recommender system which supports industries in the identification of waste items of interest in a

marketplace. The recommender system discussed in this chapter can support

industries through the identification of item opportunities in waste marketplaces, enhancing activities that may lead to the development of an active waste exchange network. To stimulate or enhance cooperation between industrial firms to utilize industrial waste streams from other industries firms share related knowledge, in order to achieve sustainable production. To build effective recommendation, we study the role of knowledge in the design of a recommender that suggests waste materials to be used in process industries. This chapter compares the performance of a knowledge based input-output recommender with a recommender based on

association rules. The two recommenders are evaluated with real-world data

collected through deploying surveys in a workshop setting. This shows that many data challenges arise when creating recommendations from explicit knowledge and suggests that techniques based on the concept of implicit knowledge may be preferable in the design of an recommender operating in an industrial symbiosis marketplace. This chapter is based on the following peer-reviewed publication:

• G. van Capelleveen, C. Amrit, D. M. Yazan and H. Zijm, ‘The influence of knowledge in the design of a recommender system to facilitate industrial

symbiosis markets’, Environmental Modelling & Software, 2018, ISSN:

1364-8152, [Online]. Available: https://doi.org/10.1016/j.envsoft. 2018.04.004

Chapter 7: The recommender canvas

This theoretical chapter describes the developed generalized framework for

recommender system design. This framework has been developed to support

the complex task of designing a recommender system. Partly driven by the

previously discussed cases of recommender design and its related literature it became noticeable that recommender design is becoming increasingly complex. This can be attributed as a result of the many technological advancements that may be included in a recommender system, that cause engineers to face a fast growing

number of design related decisions to be taken. Unfortunately, there is no general approach yet for decision makers that can act as a framework guiding the design of a recommender system. The rich collection of literature on recommender systems, though, offers a great source to identify the key areas where these decisions need to be taken. In this chapter, we survey existing literature with the aim of building a recommender system model inspired by Osterwalder’s canvas theory. The result of our semi-structured synthesis is a novel design approach in the form of a canvas for designing recommender systems. This work provides a better understanding and can serve as a guide for decision making in recommender system design. This chapter is based on the following peer-reviewed publication:

• G. van Capelleveen, C. Amrit, D. M. Yazan and H. Zijm, ‘The recommender canvas: A model for developing and documenting recommender system design’, Expert Systems with Applications, vol. 129, pp. 97–117, 2019,ISSN: 0957-4174, [Online]. Available: https://doi.org/10.1016/j.eswa. 2019.04.001

Chapter 8: Conclusions

Finally, the last chapter of this thesis concludes with revisiting the research questions, discussing recommender system design for sustainable industries, explaining the supplementary research conducted during the thesis, and providing

directions for future research. This chapter is partly based on the following

(co)-authored peer-reviewed publication:

• L. Fraccascia, G. van Capelleveen, V. Yazdanpanah and D. M. Yazan, ‘Resource inventory for fostering industrial symbiosis practices’, English, in Methods and tools for the implementation of industrial symbiosis, E.

Mancuso and V. Fantin, Eds., Italy: ENEA, Oct. 2017, pp. 51–52, ISBN:

978-88-8286-358-6

• G. van Capelleveen, J. Pohl, A. Fritsch and D. Schien, ‘The footprint

of things: A hybrid approach towards the collection, storage and

distribution of life cycle inventory data’, English, in ICT4S2018. 5th

International Conference on Information and Communication Technology for Sustainability, ser. EPiC series in computing, United Kingdom: EasyChair, May 2018, pp. 350–364, [Online]. Available: https : / / doi . org/10.29007/8pnj

• L. Fraccascia, V. Yazdanpanah, G. van Capelleveen and D. M. Yazan, ‘A framework of industrial symbiosis systems for agent based simulation’, in

1.3. ORGANIZATION OF THE THESIS 13

IEEE Conference on Business Informatics, IEEE, 2019, [Online]. Available: https://doi.org/10.1109/CBI.2019.00055

• O. Genc, G. van Capelleveen, E. Erdis, O. Yildiz and D. M. Yazan, ‘A socio-ecological approach to improve industrial zones towards eco-industrial parks’, Journal of Environmental Management, vol. 250, p. 109 507, 2019,

ISSN: 0301-4797, [Online]. Available: https://doi.org/10.1016/j.

jenvman.2019.109507

• L. Fraccascia, V. Yazdanpanah, G. van Capelleveen and D. M. Yazan, ‘Energy-based Industrial Symbiosis A literature review for circular energy transition’, Environment, Development and Sustainability, 2020 (Under review)

CHAPTER 2

A review of research gaps

The background chapter provides a scope and history of information systems that facilitate industrial symbiosis identification based on a structured literature review. It contributes a classification framework which is used to analyze information systems that act as a facilitator of communication and distributor of knowledge and describes (i) the characteristics of these different information systems, (ii) supporting role they can provide, and (iii) the technologies used to enable such identification. This overview illustrates the scientific knowledge gap we have observed with respect to the role of recommender systems for industrial symbiosis.

2.1

Introduction

In the previous chapter we have stressed the importance of industrial symbiosis as a means to contribute to the sustainability of materials in the process industry and introduced recommender systems to be considered as a technology to support

managerial work. Recommender systems can be a great addition to various

information systems facilitating the use of industrial symbiosis. However, the design of an information system is usually highly characterized by the domain of application. In order to understand how and where a recommender system can provide a functional contribution, and which system or domain characteristics may influence the system design, we would benefit from a structured review of literature regarding information systems facilitating industrial symbiosis. Although the work of Grant et al. [52] pioneers a review of systems, its preliminary assessment is based on a selected sample of waste marketplaces. A structured overview of all the different type of systems facilitating industrial symbiosis still lacks in the academic literature. It should become clear which roles information systems can play to support industrial symbiosis development and what the current state of knowledge is regarding the implementation context. This is in our observation still poorly understood and the motivation for this chapter. Therefore we aim to provide a better understanding of the characteristics of the systems, the functional support these systems can provide and the technologies used to enable identification of industrial symbiosis. To that end, we have conducted a thorough literature survey

of systems facilitating the identification of industrial symbiosis [1], which forms the basis of this chapter.

Such a study may also provide a good starting point to outline ways of progression and improvement for identification systems. The study [1] contributes a conceptual framework to practitioners and policy makers that outlines the course of actions of information systems to the goal of industrial symbioses identification. Moreover, the review reveals research directions for the industrial symbioses community to consider by suggesting advancements to be made in order to strengthen community engagement and to enrich data use for computerized exploration. The key knowledge gap revealed from this study, which inspired the central concept in this thesis, concerns the design of recommender systems for identifying industrial symbiosis in all areas of the framework, but in specific, to the promising area of facilitated synergy identification systems.

The recent popularity of industrial symbiosis has attracted and directed also other scholars to review the progress of (information) systems for the identification of industrial symbiosis with similar or slightly different discussions (compared to [1], [52]) on how these systems contribute to the identification of industrial symbiosis opportunities (see [78]–[80], [95]). Also these reviews still indicate a lack of studies on the design, implementation and evaluation of recommender

systems1 for industrial symbiosis. To the best of our knowledge, following the

definition of a recommender system defined in this thesis, the only known academic works on recommender systems for industrial symbiosis until now [2]–[5], [96], [97] have recently been developed in the SHAREBOX project [98].

2.2

Methodology

The objective of this literature review is to provide a clear overview on the different information systems that facilitate industrial symbiosis network development, the type of support they provide and the technologies being researched to establish this support. Moreover, we attempt to reveal the practical value of each type of information system and identify the challenges and research directions to strengthen the support capacity of these tools.

1_{The term ’recommender system’ or ’recommendation’ is often used with a different definition in}

industrial symbiosis than the one proposed in [1] and in this thesis. With the term recommender systems this literature points to either (i) decision support tools that apply to the earlier described ’assessment’ phase which typically relate to methods such as simulation (e.g., modeling of stock, economic costs, environmental cost, etc.) or (ii) a match-making process (e.g., waste-to-resource matching) that lacks a computer-implementable algorithm that predicts the ”rating” or ”preference” a user would attribute to an item in order to actively suggest items of interest (instead the systems often describe a manual database lookup to find a match based on the corresponding EWC code as part of a consulting procedure).

2.2. METHODOLOGY 17

The study is based on a bibliographic search conducted in February 2017. Figure 2.1 graphically shows the steps conducted in this research. Initially, we conduct a Scopus [99] search for bibliographic scientific material. We analyze the industrial symbiosis literature from 2000 till 2016. This time-frame is selected based on the publication date of the well cited paper [8], which can be seen

as one of the early foundations on the concept of industrial symbiosis. For

an extensive analysis of the literature, a broad search is required to capture the scarcely published research on industrial symbiosis tools. In particular, we must take an interdisciplinary approach in retrieving articles from a variety of journals and conferences, as the information system literature is scattered over multiple disciplinary fields ranging from ecological ecosystems to computer sciences.

Figure 2.1: Steps for the literature review.

Our queries search for a set of keywords in the title, abstract and keywords

and ”information systems” and were applied in various combinations between

the concepts. The keywords are ”industrial symbiosis, industrial ecology,

eco-industrial park, synergy, by-product exchange, recycling network, waste exchange”, and ”information system, ICT, tool, decision support, intelligence, expert system, identification, assessment, mapping”. However, some papers were excluded from the analysis because the full text was not available. We limited our search by papers published in the English language. As the result of the initial search, 351 papers were found. We used the following selection criteria to filter relevant industrial symbiosis articles that we included in our review:

• The article mentions, or presents an industrial symbiosis identification tool that is primarily aims to reveal options, industries or individual organizations in support of waste exchange or industrial symbiosis network development. • The tool is considered an information system, and thus utilizes various ICT

techniques.

• The article provides extensive information on the implementation of the tool or elaborates on the techniques that are essential and specific to this type of information systems.

These criteria can be summarized as the paper should contain a case description of at least one ICT-based IS synergy identification system. After this selection 18 papers remained that met the above criteria. To prevent missing articles due to the use of inconsistent terminology practiced over different disciplines, we perform a first degree backward and forward analysis based on the citations of the relevant publications (snowball technique). Another 19 articles were added that met the above filtering criteria. The final database, composed only by papers reporting at least one ICT-based information system facilitating industrial symbiosis, is made by 37 papers published in journals or conferences.

Next, we used the selected literature to conceptualize the industrial symbiosis systems into a matrix using the methodology of [100] to further develop an understanding of the key concepts of the information systems that stimulate

industrial symbiosis facilitation. To compose the classification we analyzed

the literature to distinguish information systems by the type of support, their characteristics and common technology and techniques used.

2.3

Industrial symbiosis tools

This section describes our typology consisting of six concepts (see Table 2.1), created from the identified literature (see Table 2.2), and summarizes advancements of each of the areas.

2.3. INDUSTRIAL SYMBIOSIS TOOLS 19

Table 2.1: Literature classification framework of information systems that facilitate industrial symbiosis identification (Note that [101], [52] and [53], [102] discuss more than one concept).

Information system Type of support Characteristics Common technology & techniques Ref. A Open online waste markets Passive facilitation of waste transactions A web-based information system that provides an open business-to-business sales platform service for waste materials. The platforms do not or only to a minimal extent interfere or coordinate transactions. Most systems form an open market and information is visible to the outside world. Typically, this type of market is mostly driven by governmental agencies or individuals without a sustainable profitable business model.

Web-based e-commerce platforms, rule-based matchmaking algorithms, ontology engineering [103] [52] [104] [105] [106] [107] [108] [81] [109] B Facilitated synergy identification systems Active facilitation of waste transactions

An information system used to support private community-based business-to-business sales of waste materials which in many cases are facilitated through an intermediary. This type of systems is used during or after identifications workshops and tend to be more structured and coordinated by the intermediary. Typically, these markets are initiated by consortia, industrial parks or third party facilitators and are primarily driven by profitability considerations. Rule-based matchmaking algorithms, national input-output tables, and life-cycle assessment databases [110] [20] [77] [111] [53] [102] [101] [52] [112] [113] [114] [115] [116] [117] [118]

Continuation of Table 2.1 C Industry sector synergy identification Profiling of waste production and use per industry

A systems approach that examines synergies between industry sectors rather than between factories. Using national waste statistics per industry sector it is possible to detect industry sectors that provide waste in large amounts or with a severe impact on the environment as well as industry sectors that may use this waste as an alternative to traditional input (possibly after some preprocessing). Using this information and the help of manufacturing process information it is possible to determine waste flows that can potentially be converted into useful process inputs, thus providing synergy between (sometimes quite different) industries. National waste input-output tables [119] [120] D Social network platforms and social network communities Mixed: a) Building social relations. b) Knowledge exchange.

Here, we refer to Social Networks (SN) that share knowledge and information on industrial symbiosis experiences and opportunities for industries. The tools unfold either as independent SN platforms or as a formation of groups on existing SN platforms.

Social networks, and social network integration [121] [122] E Industrial symbiosis knowledge repositories Knowledge exchange

Knowledge systems that contain a collection of historical industrial symbiosis examples or theoretical literature references to potential industrial symbiosis cases. The systems enable collaborative knowledge creation by providing a platform to share and discuss synergy implementation experiences. Content Management systems (e.g., Wiki’s), and collaborative web-based spreadsheets [101] [123], [124] [37] [53] [125] [126]

2.3. INDUSTRIAL SYMBIOSIS TOOLS 21 Continuation of Table 2.1 F Region identification system for industrial symbiosis Urban planning and policy making Geographical information systems (GIS) that estimate the potential of areas for the application of industrial symbiosis development GIS visualizations, Geographical-data based scoring algorithms, fuzzy rule based expert systems, and simulation software [53] [102] [127] [128] [129] [130] [131]

Table 2.2: Relevant literature classified Ref. Type of study Information system Type of waste/industry Location A B C D E F [103] Simulation Webfill Demolition

waste Hong Kong X [104] [105] Model TRXP Electronic waste United Kingdom X [106] [107] [108]

Case Study e-Symbiosis All industrial waste Greece and United Kingdom X

[81] Case study MNExchange .org All industrial waste except hazardous waste Minnesota, USA X [109] Modeling method Theoretical system description - - X

[110] Case Study WasteX All industrial waste

Jamaica X

[20] Case Study IUWA Waste Manager All industrial waste Rhein-Neckar Region, Germany X

[77] Case Study NISP information management tool All industrial waste United Kingdom X

Continuation of Table 2.2 [111] Case Study CSRP Global Synergy Database Mineral industry Kwinana, Australia X

[101] Case Study Facility Synergy Tool (EPA software) Eco-Industrial Park Rio de Janeiro, Brazil X X [52] Review of system Multiple Examples Multiple industries Multiple locations X X [112] [113]

Case Study ENEA Symbiosis All industrial waste Catania, Italy X [114] Case Study Conference Independent of

industry

Link¨oping, Sweden

X [115] Case Study SymbioSyS

tool Predominantly automotive sector Cantabria, Northern Spain X [116] [117] [118]

Case Study J-Park Simulator

All industrial waste

Singapore X

[119] Case Study Conceptual system description with data Fossil fuel, metal and mineral waste; agricultural and synthetic waste; electronic waste Taiwan X [120] Framework development Theoretical system description - - X [121] Framework development Theoretical system description - - X [122] System Reference LinkedIn Industrial Symbiosis - - X [123] [124]

Case Study Enipedia All industrial waste

2.3. INDUSTRIAL SYMBIOSIS TOOLS 23 Continuation of Table 2.2 [37] Reference Case Study ISData.org (industrial symbiosis4IE) All industrial waste Worldwide X [125] Reference Case Study Collection of NISP case studies All industrial waste Worldwide X [126] Reference Case Study Nordregio industrial symbiosis cases All industrial waste Worldwide X [53] [102]

Case Study The Geneva Industrial Symbiosis Project (used Presteo) (including the Industrie et Synergies Inter-Sectorielles database) All industrial waste Geneva, Switzerland X X X

[127] Case Study INGEPRO Research Group: Spatial Decision Support Systems All industrial waste Cantabria, Northern Spain X

[131] Case Study Habitat suitability Index (NISP) All industrial waste South Hampton, UK X

[128] Case Study Looplocal All industrial waste Sweden X [129] [130] Model framework, Case Study Tool for potential of a District Heating System (DHS)

Heat waste Fukushima region, Japan