Nicola Ortelli

Journal publications [3]

2024

Ortelli N., de Lapparent M., Bierlaire M. (2024). Resampling Estimation of Discrete Choice Models. Journal of Choice Modelling 50 (100467)

In the context of discrete choice modeling, the extraction of potential behavioral insights from large datasets is often limited by the poor scalability of maximum likelihood estimation. This paper proposes a simple and fast dataset-reduction method that is specifically designed to preserve the richness of observations originally present in a dataset, while reducing the computational complexity of the estimation process. Our approach, called LSH-DR, leverages locality-sensitive hashing to create homogeneous clusters, from which representative observations are then sampled and weighted. We demonstrate the efficacy of our approach by applying it on a real-world mode choice dataset: the obtained results show that the samples generated by LSH-DR allow for substantial savings in estimation time while preserving estimation efficiency at little cost.
```
@Article{OrteLappBier2024,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Resampling Estimation of Discrete Choice Models},
  journal = {Journal of Choice Modelling},
  year = {2024},
  volume = {50},
  number = {100467},
  DOI = {10.1016/j.jocm.2023.100467},
  note = {Accepted on Dec 20, 2023}}}
```

2021

Ortelli N., Hillel T., Pereira F., de Lapparent M., Bierlaire M. (2021). Assisted Specification of Discrete Choice Models. Journal of Choice Modelling 39 (100285)

Determining appropriate utility specifications for discrete choice models is time-consuming and prone to errors. As the number of possible functions exponentially grows with the vari- ables under consideration, analysts need to spend increasing amounts of time on searching for good specifications through trial-and-error and expert knowledge. This paper proposes a data-driven algorithm that aims at assisting modelers in their search. Our approach trans- lates the task into a multi-objective combinatorial optimization problem and makes use of a variant of the variable neighborhood search algorithm to generate solutions. We apply the algorithm to a real mode choice dataset as a proof of concept. The results demonstrate its ability to generate high-quality specifications in reasonable amounts of time.
```
@Article{OrteHillCamadeLaBier21,
  author = {Nicola Ortelli and Tim Hillel and Francisco Pereira and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Assisted Specification of Discrete Choice Models},
  journal = {Journal of Choice Modelling},
  year = {2021},
  volume = {39},
  number = {100285},
  DOI = {10.1016/j.jocm.2021.100285},
  note = {Accepted on Mar 16, 2021}}}
```

2020

Rodrigues F., Ortelli N., Bierlaire M., Pereira F. (2020). Bayesian Automatic Relevance Determination for Utility Function Specification in Discrete Choice Models. IEEE Transactions on Intelligent Transportation Systems


@Article{RodrOrteBierPere20,
  author = {Filipe Rodrigues and Nicola Ortelli and Michel Bierlaire and Francisco Pereira},
  title = {Bayesian Automatic Relevance Determination for Utility Function Specification in Discrete Choice Models},
  journal = {IEEE Transactions on Intelligent Transportation Systems},
  year = {2020},
  DOI = {10.1109/TITS.2020.3031965},
  note = {Accepted on Oct 14, 2020}}}

Technical reports [5]

2024

Ortelli N., de Lapparent M., Varotto S., Bierlaire M. (2024). An integrated ordered logit and latent variable model for accident injury severity and risk-taking behavior. Report TRANSP-OR 240101


@TechReport{OrteLappVaroBier2024,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Silvia Varotto and Michel Bierlaire},
  title = {An integrated ordered logit and latent variable model for accident injury severity and risk-taking behavior},
  institution = {Transport and Mobility Laboratory, Ecole Polytechnique F\'ed\'erale de Lausanne},
  year = {2024},
  type = {Technical Report},
  number = {TRANSP-OR 240101},
  address = {Lausanne, Switzerland}}}

2023

Ortelli N., de Lapparent M., Bierlaire M. (2023). Resampling estimation of discrete choice models. Report TRANSP-OR 230330


@TechReport{OrteLappBier2023,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Resampling estimation of discrete choice models},
  institution = {Transport and Mobility Laboratory, Ecole Polytechnique F\'ed\'erale de Lausanne},
  year = {2023},
  type = {Technical Report},
  number = {TRANSP-OR 230330},
  address = {Lausanne, Switzerland}}}

Bierlaire M., Ortelli N. (2023). Assisted specification with Biogeme 3.2.12. Report TRANSP-OR 230816


@TechReport{BierOrte23,
  author = {Michel Bierlaire and Nicola Ortelli},
  title = {Assisted specification with Biogeme 3.2.12},
  institution = {Transport and Mobility Laboratory, Ecole Polytechnique F\'ed\'erale de Lausanne},
  year = {2023},
  type = {Technical Report},
  number = {TRANSP-OR 230816},
  address = {Lausanne, Switzerland}}}

2022

Bierlaire M., Ortelli N. (2022). Assisted specification with Biogeme. Report TRANSP-OR 220707


@TechReport{BierOrte22,
  author = {Michel Bierlaire and Nicola Ortelli},
  title = {Assisted specification with Biogeme},
  institution = {Transport and Mobility Laboratory, Ecole Polytechnique F\'ed\'erale de Lausanne},
  year = {2022},
  type = {Technical Report},
  number = {TRANSP-OR 220707},
  address = {Lausanne, Switzerland}}}

2020

Ortelli N., Hillel T., Pereira F., de Lapparent M., Bierlaire M. (2020). Assisted Specification of Discrete Choice Models. Report TRANSP-OR 200708


@TechReport{OrtHilPerLapBie2020,
  author = {Nicola Ortelli and Tim Hillel and Francisco Pereira and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Assisted Specification of Discrete Choice Models},
  institution = {Transport and Mobility Laboratory, Ecole Polytechnique F\'ed\'erale de Lausanne},
  year = {2020},
  type = {Technical Report},
  number = {TRANSP-OR 200708},
  address = {Lausanne, Switzerland}}}

Articles in conference proceedings [7]

2024

Ortelli N., de Lapparent M., Varotto S., Bierlaire M. (2024) An integrated accident injury severity and risk-taking behavior model for the evaluation of Via Sicura. Proceedings of the 12th Symposium of the European Association for Research in Transportation (HEART), 18-20 June 2024, Espoo, Finland


@InProceedings{OrtHilPerLapBie_HEART2024,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Silvia Varotto and Michel Bierlaire},
  title = {An integrated accident injury severity and risk-taking behavior model for the evaluation of Via Sicura},
  booktime = {Proceedings of the 12th Symposium of the European Association for Research in Transportation (HEART), 18-20 June 2024},
  year = {2024},
  address = {Espoo, Finland}}}

Ortelli N., de Lapparent M., Bierlaire M. (2024) A conditional trust-region algorithm for the estimation of discrete choice models. Proceedings of the 24th Swiss Transport Research Conference (STRC), 15-17 May, Ascona, Switzerland


@InProceedings{OrtLapBie_STRC2024,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {A conditional trust-region algorithm for the estimation of discrete choice models},
  booktime = {Proceedings of the 24th Swiss Transport Research Conference (STRC), 15-17 May},
  year = {2024},
  address = {Ascona, Switzerland}}}

2023

Ortelli N., de Lapparent M., Bierlaire M. (2023) Faster estimation of discrete choice models via weighted dataset reduction. Proceedings of the 11th Symposium of the European Association for Research in Transportation (HEART), 6-8 September 2023, Zurich, Switzerland


@InProceedings{OrtHilPerLapBie_HEART2023,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Faster estimation of discrete choice models via

weighted dataset reduction},
  booktime = {Proceedings of the 11th Symposium of the European Association for Research in Transportation (HEART), 6-8 September 2023},
  year = {2023},
  address = {Zurich, Switzerland}}}

Ortelli N., de Lapparent M., Bierlaire M. (2023) Stochastic adaptive resampling for the estimation of discrete choice models. Proceedings of the 23rd Swiss Transport Research Conference (STRC), 10-12 May, Ascona, Switzerland


@InProceedings{OrtLapBie_STRC2023,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Stochastic adaptive resampling for the estimation of discrete choice models},
  booktime = {Proceedings of the 23rd Swiss Transport Research Conference (STRC), 10-12 May},
  year = {2023},
  address = {Ascona, Switzerland}}}

2022

Ortelli N., de Lapparent M., Bierlaire M. (2022) Faster estimation of discrete choice models via dataset reduction. Proceedings of the 22nd Swiss Transport Research Conference (STRC), 18-20 May, Ascona, Switzerland


@InProceedings{OrtLapBie_STRC2022,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Faster estimation of discrete choice models via dataset reduction},
  booktime = {Proceedings of the 22nd Swiss Transport Research Conference (STRC), 18-20 May},
  year = {2022},
  address = {Ascona, Switzerland}}}

2021

Ortelli N., Bierlaire M., de Lapparent M. (2021) Can we infer on behavioral impacts of public policy on accident severity outcomes? A Swiss case study using historical disaggregate accident reports. Proceedings of the 21st Swiss Transport Research Conference (STRC), 12-14 September, Ascona, Switzerland


@InProceedings{OrtBieLap_STRC2021,
  author = {Nicola Ortelli and Michel Bierlaire and Matthieu {de Lapparent}},
  title = {Can we infer on behavioral impacts of public policy on accident severity outcomes? A Swiss case study using historical disaggregate accident reports},
  booktime = {Proceedings of the 21st Swiss Transport Research Conference (STRC), 12-14 September},
  year = {2021},
  address = {Ascona, Switzerland}}}

2020

Ortelli N., Hillel T., Pereira F., de Lapparent M., Bierlaire M. (2020) Variable Neighborhood Search for Assisted Utility Specification in Discrete Choice Models. Proceedings of the 9th Symposium of the European Association for Research in Transportation (HEART), 3-4 February 2021, Lyon, France


@InProceedings{OrtHilPerLapBie_HEART2020,
  author = {Nicola Ortelli and Tim Hillel and Francisco Pereira and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Variable Neighborhood Search for Assisted Utility Specification in Discrete Choice Models},
  booktime = {Proceedings of the 9th Symposium of the European Association for Research in Transportation (HEART), 3-4 February 2021},
  year = {2020},
  address = {Lyon, France}}}

Talks [11]

2024

Ortelli N., de Lapparent M., Varotto S., Bierlaire M. An integrated injury severity and risk-taking behavior model for the evaluation of Via Sicura June 19, 2024. 12th Symposium of the European Association for Research in Transportation, Aalto University, Espoo, Finland.


@Misc{Ortelli_hEART2024,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Silvia Varotto and Michel Bierlaire},
  title = {An integrated injury severity and risk-taking behavior model for the evaluation of Via Sicura},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_hEART2024.pdf}},
  year = {2024},
  note = {Presented at 12th Symposium of the European Association for Research in Transportation, Aalto University, Espoo, Finland}}}

Ortelli N., de Lapparent M., Bierlaire M. A conditional trust-region algorithm for the estimation of discrete choice models May 16, 2024. 24th Swiss Transport Research Conference, Monte Verità, Ascona, Switzerland.


@Misc{Ortelli_strc2024,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {A conditional trust-region algorithm for the estimation of discrete choice models},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_strc2024.pdf}},
  year = {2024},
  note = {Presented at 24th Swiss Transport Research Conference, Monte Verit\à, Ascona, Switzerland}}}

2023

Ortelli N., Varotto S., Bierlaire M., de Lapparent M. Un nouveau modèle hybride pour la modélisation de la conduite imprudente et de la gravité des accidents October 12, 2023. Séminaire interne de l'IIDE, Yverdon-les-Bains, Switzerland.


@Misc{Ortelli_iide2023,
  author = {Nicola Ortelli and Silvia Varotto and Michel Bierlaire and Matthieu {de Lapparent}},
  title = {Un nouveau mod\èle hybride pour la mod\élisation de la conduite imprudente et de la gravit\é des accidents

},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_iide2023.pdf}},
  year = {2023},
  note = {Presented at S\éminaire interne de l'IIDE, Yverdon-les-Bains, Switzerland}}}

Ortelli N., de Lapparent M., Bierlaire M. Resampling estimation of discrete choice models September 07, 2023. 11th Symposium of the European Association for Research in Transportation, ETHZ, Zurich, Switzerland.


@Misc{Ortelli_hEART2023,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Resampling estimation of discrete choice models

},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_hEART2023.pdf}},
  year = {2023},
  note = {Presented at 11th Symposium of the European Association for Research in Transportation, ETHZ, Zurich, Switzerland}}}

Ortelli N., de Lapparent M., Bierlaire M. Stochastic adaptive resampling for the estimation of discrete choice models June 01, 2023. 15th workshop on Discrete Choice Models, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland.


@Misc{Ortelli_dcm2023,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Stochastic adaptive resampling

for the estimation of discrete choice models},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_dcm2023.pdf}},
  year = {2023},
  note = {Presented at 15th workshop on Discrete Choice Models, Ecole Polytechnique F\éd\érale de Lausanne, Lausanne, Switzerland}}}

Ortelli N., de Lapparent M., Bierlaire M. Stochastic adaptive resampling for the estimation of discrete choice models May 11, 2023. 23rd Swiss Transport Research Conference, Monte Verità, Ascona, Switzerland.


@Misc{Ortelli_strc2023,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Stochastic adaptive resampling for the estimation of discrete choice models},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_strc2023.pdf}},
  year = {2023},
  note = {Presented at 23rd Swiss Transport Research Conference, Monte Verit\à, Ascona, Switzerland}}}

Ortelli N., Varotto S., Bierlaire M., de Lapparent M. An integrated ordered logit and latent variable model for crash injury severity: a Swiss case study using disaggregate crash reports February 16, 2023. 1st EPFL Symposium on Transportation Research, EPFL, Annecy, France.


@Misc{Ortelli_sympo2023,
  author = {Nicola Ortelli and Silvia Varotto and Michel Bierlaire and Matthieu {de Lapparent}},
  title = {An integrated ordered logit and latent variable model for crash injury severity: a Swiss case study using disaggregate crash reports},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_sympo2023.pdf}},
  year = {2023},
  note = {Presented at 1st EPFL Symposium on Transportation Research, EPFL, Annecy, France}}}

2022

Ortelli N., Varotto S., Bierlaire M., de Lapparent M. An integrated ordered logit and latent variable model for crash injury severity: a Swiss case study using disaggregate crash reports December 15, 2022. 16th International Conference on Travel Behaviour Research, International Association for Travel Behaviour Research (IATBR), Santiago, Chile.


@Misc{Ortelli_iatbr2022,
  author = {Nicola Ortelli and Silvia Varotto and Michel Bierlaire and Matthieu {de Lapparent}},
  title = {An integrated ordered logit and latent variable model for crash injury severity: a Swiss case study using disaggregate crash reports},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_iatbr2022.pdf}},
  year = {2022},
  note = {Presented at 16th International Conference on Travel Behaviour Research, International Association for Travel Behaviour Research (IATBR), Santiago, Chile}}}

Ortelli N., de Lapparent M., Bierlaire M. Faster estimation of discrete choice models via dataset reduction May 19, 2022. 22nd Swiss Transport Research Conference, Monte Verità, Ascona, Switzerland.


@Misc{Ortelli_strc2022,
  author = {Nicola Ortelli and Matthieu {de Lapparent} and Michel Bierlaire},
  title = {Faster estimation of discrete choice models via dataset reduction},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_strc2022.pdf}},
  year = {2022},
  note = {Presented at 22nd Swiss Transport Research Conference, Monte Verit\à, Ascona, Switzerland}}}

2021

Ortelli N., Bierlaire M., de Lapparent M. Can we infer on behavioral impacts of public policy on accident severity outcomes? September 13, 2021. 21st Swiss Transport Research Conference, Monte Verità, Ascona, Switzerland.


@Misc{Ortelli_strc2021,
  author = {Nicola Ortelli and Michel Bierlaire and Matthieu {de Lapparent}},
  title = {Can we infer on behavioral impacts of public policy on accident severity outcomes?},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_strc2021.pdf}},
  year = {2021},
  note = {Presented at 21st Swiss Transport Research Conference, Monte Verit\à, Ascona, Switzerland}}}

Ortelli N. Assisted specification of discrete choice models July 13, 2021. CMC Online Seminar Series 2021, Choice Modelling Centre (CMC), University of Leeds, Leeds, UK.


@Misc{Ortelli_leeds2021,
  author = {Nicola Ortelli},
  title = {Assisted specification of discrete choice models},
  howpublished = {\url{https://transp-or.epfl.ch/documents/talks/Ortelli_leeds2021.pdf}},
  year = {2021},
  note = {Presented at CMC Online Seminar Series 2021, Choice Modelling Centre (CMC), University of Leeds, Leeds, UK}}}

Reviewing [2]

EPFL courses [5]

Fall 2023

Mathematical modeling of behavior.
- Level: Master
- Sections: Mathematics, Master in Financial Engineering
- Lecturer: Michel Bierlaire
- Teaching assistants: Evangelos Paschalidis, Nicola Ortelli, Janody Pougala, Cloe Cortes Balcells, Pavel Ilinov, Risa Kobayashi
- Webpage: https://moodle.epfl.ch/course/view.php?id=1001

Fall 2022

Mathematical modeling of behavior.
- Level: Master
- Sections: Mathematics, Master in Financial Engineering
- Lecturer: Michel Bierlaire
- Teaching assistants: Silvia Varotto, Nicola Ortelli, Cloe Cortes Balcells, Tom Haering
- Webpage: https://moodle.epfl.ch/course/view.php?id=1001

Fall 2021

Mathematical modeling of behavior.
- Level: Master
- Sections: Mathematics, Master in Financial Engineering
- Lecturer: Virginie Lurkin
- Teaching assistants: Melvin Wong, Silvia Varotto, Nicola Ortelli, Janody Pougala, Rico Krueger, Tim Hillel
- Webpage: https://moodle.epfl.ch/course/view.php?id=1001

Fall 2020

Mathematical modeling of behavior.
- Level: Master
- Sections: Mathematics, Master in Financial Engineering
- Lecturer: Michel Bierlaire
- Teaching assistants: Melvin Wong, Rico Krueger, Janody Pougala, Nicola Ortelli, Meritxell Pacheco, Tim Hillel
- Webpage: https://moodle.epfl.ch/course/view.php?id=1001

Fall 2019

Mathematical modeling of behavior.
- Level: Master
- Sections: Mathematics, Master in Financial Engineering
- Lecturer: Michel Bierlaire
- Teaching assistants: Tim Hillel, Meritxell Pacheco, Janody Pougala, Nicola Ortelli
- Webpage: https://moodle.epfl.ch/course/view.php?id=1001

Lectures [6]

2024

Logit Mixtures; Combining Data (SP and RP).
- Discrete Choice Analysis: Predicting Individual Behavior and Market Demand, EPFL, Lausanne, Switzerland
- Feb 13, 2024
- Lecturer: Evangelos Paschalidis
- Assistants: Cloe Cortes Balcells, Nicola Ortelli
Introduction, Logit Estimation and Testing.
- Discrete Choice Analysis: Predicting Individual Behavior and Market Demand, EPFL, Lausanne, Switzerland
- Feb 11, 2024
- Lecturer: Tom Haering
- Assistants: Negar Rezvany, Nicola Ortelli, Fabian Torres

2023

Hierarchical Bayesian Estimation; Individual Prediction.
- Discrete Choice Analysis: Predicting Individual Behavior and Market Demand, EPFL, Lausanne, Switzerland
- Feb 02, 2023
- Lecturer: Janody Pougala
- Assistants: Marija Kukic, Nicola Ortelli
Hybrid choice models.
- Discrete Choice Analysis: Predicting Individual Behavior and Market Demand, EPFL, Lausanne, Switzerland
- Feb 01, 2023
- Lecturer: Nicola Ortelli
- Assistants: Tom Haering, Evangelos Paschalidis, Janody Pougala
Logit Estimation and Specification Testing.
- Discrete Choice Analysis: Predicting Individual Behavior and Market Demand, EPFL, Lausanne, Switzerland
- Jan 29, 2023
- Lecturer: Tom Haering
- Assistants: Cloe Cortes Balcells, Marija Kukic, Nicola Ortelli

2021

Hybrid Choice Models.
- Discrete Choice Analysis: Predicting Individual Behavior and Market Demand, EPFL, Lausanne, Switzerland
- Feb 12, 2021
- Lecturer: Nicola Ortelli
- Assistant: Rico Krueger

MSc theses [2]

2022

Stochastic and robust hydropower unit commitment, Pierre Tarazi (Electrical and Electronics Engineering)
- Supervision: Tom Haering, Nicola Ortelli
- External expert: Maziar Kermani

2020

Feuille de route pour l'électrification du réseau de bus urbains de Lausanne à l'horizon 2030, Alain Azzi (Mechanical Engineering)
- Supervision: Nicola Ortelli, Janody Pougala, Michel Bierlaire
- External expert: Philippe Seydoux

Semester projects [10]

2024

Improving the specification of migration aspiration discrete choice models via attitudinal variables, Anne-Valérie Preto (Civil Engineering)

Supervision: Evangelos Paschalidis, Nicola Ortelli, Michel Bierlaire
International migration is a crucial but controversial topic that raises issues and concerns to be resolved in the parties involved. In industrialised nations, the proportion of immigrants in total population increased between 1960 and 2019, raising fears about economic costs for natives, loss of national identity, and integra- tion issues. In poor countries, international migration raises concerns regarding brain drain of highly-skilled people, as college and university graduates have a higher tendency to emigrate than the less educated (Beine et al., 2021). The accurate understanding and prediction of migration motifs are crucial from a policy- making perspective, to measure for instance the immigration pressure a country faces. Improved models can help to predict and adjust migration flows but also to identify and mitigate adverse effects on all parties involved. The objective of this project is to extend advanced discrete choice model specifications that address issues such as the independence from irrelevant alternatives (IIA). For instance, Beine et al. (2021) implemented a cross-nested logit (CNL) model to approximate the substitution patterns in migration aspirations of Indian individuals more accurately. The specification of such models could be improved by incorporating addi- tional variables related to risk aversion of the individuals, perceived security of their environment, perceived degradation of the climatic conditions or perceived governance in their country could influence aspiration. To address issues related to measurement errors, these variables will be incorporated in the form of latent variables which have been extensively used in the past in the field of discrete choice models. Ultimately, the objective of this project is to extend the use of latent variables in the context of migration aspiration models and examine whether the insights of previous models (such as the CNL), for instance in terms of substitution, hold when improving this specification.

2023

A detailed evaluation of the Via Sicura program, Mya Jamal Lahjouji (Civil Engineering)

Supervision: Nicola Ortelli, Selin Atac, Michel Bierlaire
This project aims at updating and improving the official evaluation of the Via Sicura road safety program.

2022-2023

Scenario generation for the multi-agent transport simulation toolkit, MATSim, Anne-Valerie Preto (Civil Engineering)

Supervision: Selin Atac, Nicola Ortelli, Michel Bierlaire
CSSs are becoming increasingly popular, primarily due to their financial and environmental advantages. However, they face many operational challenges, including inventory management of vehicles and parking spots, vehicle load balancing and redistribution, pricing strategies, and demand forecasting. If these challenges are not addressed properly, the CSS risks experiencing a significant loss of customers and therefore revenue. The literature consists of variety of works on CSSs especially focused on the rebalancing operations. On the other hand, it lacks the detailed comparison of different strategies of rebalancing operations at a disaggregate level. One approach to fill this gap is using an agent-based transport simulation. Such simulation toolkits allow analysis at the individual level and testing different configurations such as city and network structure. Therefore, in this study the student will review the literature to determine the main characteristics of the CSSs. In the first part of the project, she will work on creating representative scenarios for the multi-agent transport simulation toolkit, MATSim, based on the findings in the literature. These scenarios will be used to compare different rebalancing operations strategies where an optimization module is used. The second part of the project will be implementing analysis scripts that will provide useful information on the simulation results.

Comparison of machine learning techniques for synthetic activity generation, Xinling Li (Civil Engineering)

Supervision: Marija Kukic, Nicola Ortelli, Michel Bierlaire
This project aims to investigate the good and the bad sides of state-of-the-art Machine Learning techniques applied in the context of synthetic population generation, focusing on the generation of attributes that de- scribe synthetic activities. Recently, there has been an increase in research focus on the application of deep generative learning algorithms for generating synthetic populations and travel behaviour modelling, including Variational Auto-Encoders (VAEs) and Generative Adversarial Networks (GANs) (Borysov et al., 2019, Wong and Farooq, 2020). Whilst initial experiments show promising results, there are still significant drawbacks with using generative modelling on population synthesis, which includes vanishing gradients, training instability or mode collapse, nor it allows for exact evaluation of the probability density of new points. Based on the literature review, we try to propose possible improvements for existing methodologies by identifying if there is a way to capture better the correlations between variables. One of the techniques that could be improved is Generative Adversarial Networks (GANs). For instance, DATGAN (Lederrey et al., 2022) is a GAN architecture that tries to give more control to the modellers when generating synthetic tabular data. It uses a Directed Acyclic Graph (DAG) to represent the correlation between variables in the dataset. This DAG is then used to define the structure of the Generator. Each variable is represented by an LSTM cell. However, due to the linear nature of LSTM cells, it is required to have a DAG instead of an acyclic graph. This directionality can be an issue when designing the DAG since it represents causality between variables. Thus, we could improve the DATGAN by removing the directionality and only specifying correlations between variables. Graph Neural Networks (GNNs) are a specific type of NN designed to mimic any type of graph, and they can be used to design the architecture for the Generator. This change would give more freedom to the user and make the method more flexible since DATGAN would not require any specific type of graph

Modelling migration intentions worldwide, Hilda Abigél Horváth (Mathematics)

Supervision: Silvia Varotto, Nicola Ortelli, Michel Bierlaire
Migration across different countries is a major topic of policy debates worldwide that receives relatively little attention in the travel behaviour literature. In industrialized countries, the proportion of immigrants in the total population has increased in recent years resulting in concerns about integration, loss of national identity and economic expenses. In low-opportunity countries, international migration raises fears of loss of highly skilled workers, who are the group most likely to emigrate. Therefore, understanding the main factors that influence the location where migrants choose to settle is of primary interest for both researchers and policy makers. Location choices might be influenced by changes in the attractiveness of the country of origin, of the country of desstination, and of the alternative destinations. A recent study developed a cross-nested logit model to predict migration intentions in India based on characteristics of the individual, characteristics of the destination country and unobserved factors shared between similar destinations. However, there is still a dearth in understanding the main factors influencing migration intentions at a global scale, in countries that differ in terms of size, level of development and willingness to emigrate. The objective of this study is to assess the transferability of the previous model predicting migration intentions in India to other countries in the world. The study will identify the main characteristics of the countries of origin that have an impact on the migration intentions. The data were extracted from the Gallup World Poll (GWP) Survey. The GWP measures a series of key indicators including migration intentions and is one of the most comprehensive datasets available worldwide. Other key indicators measured in the survey are national security, nutrition, housing, job opportunities, financial situation, personal health, civil engagement and overall well-being. The GWP surveys cover more than 150 countries in the world, are usually repeated every year, and include 1000 individuals in each country. In this study, the GWP data will be analysed to select relevant countries based on the coverage, the number of waves available, the respondents per wave, and the variables available. The analysis will identify patterns per country and evolution of migration intentions over time. Mathematical methods will be implemented to understand which variables should be included to improve the model specification and to capture the perceptions of respondents about the living conditions in their country. These methods will be also used to explore the complex correlation patterns between the attributes and the destination choices.

2022

Comparing rebalancing operations in car sharing systems, Xinling Li (Civil Engineering)

Supervision: Selin Atac, Nicola Ortelli, Michel Bierlaire
A car sharing system (CSS) offers users to rent cars for a short period of time. The users are identified by an RFID card or through a mobile application. The price of the trip is generally determined according to trip duration and length. One of the challenges faced in such systems is the imbalance of the vehicles throughout the city. This problem is solved by forecasting the demand and rebalancing the vehicles to meet user demand. First, the student is going to review the literature related to rebalancing operations in CSSs. Then, she will design and/or implement different rebalancing strategies in order to test the effects of different characteristics of the city/region. In other words, different strategies (static, reactive, and proactive) for rebalancing operations will be tested for each scenario and the results will be used to drive to conclusions which explain the relation between the city structure, demand structure and the different strategies of rebalancing operations.

2021

How to assess synthetic data?, Romain Palazzo (Mathematics)

Supervision: Gael Lederrey, Nicola Ortelli, Michel Bierlaire

Road Safety: Investigating the Role of Latent Behaviors in Injury Severity, Daniela Spadaro (Mathematics)

Supervision: Nicola Ortelli, Gael Lederrey, Michel Bierlaire
Reducing the number of fatalities caused by traffic accidents is a major concern everywhere around the world. In order to do so, it is crucial to understand how various factors influence injury severity. The existing literature generally deals with explaining the influence of directly observable variables on the occurence and severity of crashes, but fails to adequately consider driver behavior. Hence, the main goal of this project is to explore approaches that enable the incorporation of latent attitudes in injury severity models. Specifically, we are interested in investigating the "propensity to taking risks" that drivers may exhibit, as well as the influence of such behaviors on the severity of injuries suffered by other road users.

2020

Enhancing the Serial Estimation of Discrete Choice Models Sequences, Youssef Kitane (Civil Engineering)

Supervision: Nicola Ortelli, Gael Lederrey, Michel Bierlaire
In the past years, an increasing number of studies has investigated the usefulness of data-driven methods for the task of developing discrete choice models so as to alleviate modelers from the burden of manual specification. One of these methods makes use of a variable neighborhood search algorithm to mimic the way an experienced modeler would build the utility functions of a discrete choice model. More precisely, small modifications are sequentially brought to an initial specification, while the induced improvement is assessed by means of a measure of performance. Although the quality of the generated models is high, this method faces a severe limitation: the computational time for a single run counts in hours with relatively small datasets and grows quickly with the amount of considered data. The main goal of this project is to speed up the estimation of DCMs in this precise context.

2019

Estimation of Discrete Choice Models using HAMABS 2.0, Linah Charif (Civil Engineering)

Supervision: Gael Lederrey, Nicola Ortelli, Michel Bierlaire
Improvement of HAMABS for choice modeling estimation using automatic switch.