Towards Efficient Rail Transportation: Bayesian Network Modeling for Predicting Passenger Train Delays Using Secondary Train Information

This paper presents the usage of a Bayesian network to predict the delay of a given passenger train on the Belgian rail network, trained with a year of arrival and departure records. In order to improve the prediction of future delays, this paper explores the possibility of incorporating the delays of other trains that visit the same station. While the biggest improvement is to be expected from looking at the train’s own delay in previous stations, there’s a measurable improvement in the prediction accuracy when secondary trains are taken into account.