Statistical learning for train delays and influence of winter climate and atmospheric icing

TL;DR

This paper applies advanced statistical learning models to analyze how winter climate and atmospheric icing affect train delays in northern Sweden, providing a data-driven understanding of delay factors and model performance.

Contribution

It introduces novel inhomogeneous Markov chain and stratified Cox models to quantify climate impacts on train delays, incorporating weather and operational variables.

Findings

Weather variables significantly influence train delays

The inhomogeneous Markov chain model achieved an average MAE of 0.088

Approximately 9% of trains may be misclassified regarding delays

Abstract

This study investigated the climate effect under consecutive winters on the arrival delay of high-speed passenger trains in northern Sweden. Novel statistical learning approaches, including inhomogeneous Markov chain model and stratified Cox model, were adopted to account for the time-varying risks of train delays. The inhomogeneous Markov chain modelling for the arrival delays has used several covariates, including weather variables, train operational direction, and findings from the primary delay analysis through stratified Cox model. The results showed that the weather variables, such as temperature, snow depth, ice/snow precipitation, and train operational direction, significantly impact the arrival delay. The performance of the fitted inhomogeneous Markov chain model was evaluated by the walk-forward validation method. The averaged mean absolute errors between the expected rates and the observed rates of the arrival delay over the train line was obtained at the level of 0.088, which implies that approximately 9% of trains may be misclassified as having arrival delays by the fitted model at a measuring point on the train line.