Analyzing and predicting SNCB’s train delays using open data

Details

Supervisors

Schaus, Pierre

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] en sciences informatiques, à finalité spécialisée

Abstract

SNCB, the Belgian public railway company, manages all train transportation in Belgium. The Belgian railway network, comprising nearly 600 stations, accommodates thousands of journeys daily. Unfortunately, delays often occur within this extensive network. This work aims to better understand the frequency of these delays and their impact on the network and to develop models for short-term delay prediction. Two architectures are evaluated for the prediction task: a Markov chain-based architecture, where models replace transition matrices, and a Markov random fields architecture, which attempts to capture the intricacies of the entire network. Metrics were developed to pinpoint stations with the greatest impact on the network, both as sources and catalysts of delays. Our results show that the Markov chain architecture, using Random Forests as models, can predict delays relatively well up to five stations ahead.