Kevin joined IRT SystemX in 2017 for his PhD work focusing on the prediction and detection of contextual anomalies on time series in a dynamic context, in relation to the analysis of public transport ridership. He looks back on his doctorate completed within the project on Augmented Traveller Information (Information Voyageurs Augmentée) and supervised by the Gustave Eiffel University.
What was the topic of your dissertation ?
My PhD focused on the analysis of data related to mobility and more specifically to public transport. The SNCF, a partner in the IVA project, provided us with a set of data on the load in trains and on the transport supply (schedules and theoretical and actual operations of trains). Initially, our objective was to determine our capacity to predict the load in transport in order to feed passenger information tools that would, for example, alert a user to a possible overload in a particular train.
The second application of my thesis then came quite naturally. My work consisted in supporting transport regulation operators so that they could better understand what was happening on the network. To do this, I developed a solution to analyse the data and to quantify how abnormal a value may become. The aim is to better understand the impact of a disruption on the network in order to better deal with it, by observing, for example, that a technical failure on a line will generate modal shifts leading to overcrowding on certain trains and undercrowding in other places.
What do you remember from your dissertation ?
After obtaining my engineering degree, IRT SystemX enabled me to improve my technical skills. For example, I learnt a lot in the field of statistics. Then, my thesis enabled me to acquire scientific rigour and to discover the world of research. In order to obtain reliable solutions to problems, it is essential to adopt a very meticulous approach, which my PhD helped me to develop.
Can you tell us about your favorite memory from your time with SystemX ?
We organised a barbecue with the IVA and MSM (Mobility Solutions Modelling) project teams. It was a bit of an offbeat moment on the École Polytechnique campus between engineers and researchers. The atmosphere was very pleasant one.
What are your plans for the future ?
I have been hired by the IRT SystemX as a research engineer as part of an ambitious program that it runs and coordinates in the area of trusted artificial intelligence. In particular, I will be looking at how AI algorithms work, and more specifically how to guarantee the trust and interpretability of their decisions.
More information about Kévin :
Name of PhD dissertation : Forecast and anomaly detection on time series with dynamic context. Application to the mining of transit ridership data.
R&D project : Augmented Traveller Information (Information Voyageurs Augmentée / IVA)
Name of School : Gustave Eiffel University (formerly Ifsttar)
Last qualification obtained before starting on the PhD : Engineering school diploma from the National Graduate School of Informatics for Industry and Enterprise (ENSIIE / École nationale supérieure d’informatique pour l’industrie et l’entreprise)) in combination with a Master’s degree in AIC research (Learning, Information, Content) from the University of Paris-Saclay.