Tree based models struggle to predict future train delays reliably

Predicting Delayed Train Trajectories on the Dutch Railway Network: Explainable AI Evaluation of Topological, Operational and Weather Features with Tree Based Ensemble Methods

Machine Learning

Summary

Predicting when trains will be delayed is important for managing railways. The authors studied the Dutch railway system and used computer models that are easy to understand to predict delays using lots of information like weather, train schedules, and track details. They found that these models can predict delays well when looking at data from the same time period, but their accuracy gets worse when trying to predict delays in the future. The drop in performance is linked to changing weather conditions and how delays are measured. The authors suggest future models need to consider seasonal changes and operational limits to improve long-term predictions.

What this means in practice

  • For railway operation teams: Improve short-term train delay predictions by integrating topological and weather data using interpretable tree-based models.
  • For transportation planners: Plan for the limitations of long-term delay forecasts by understanding the role of environmental volatility on prediction accuracy.

Tested on one dataset.

Authors

Jia Long Bao, Ali Mohammed Mansoor Alsahag, Seyed Sahand Mohammadi Ziabari

Abstract

The reliable prediction of passenger train delays is a critical component of railway management. While contemporary research frequently attempts to maximize absolute accuracy by deploying opaque deep learning architectures, the underlying data mechanics driving longitudinal predictive decay remain underexplored. Consequently, this study provides an explainable temporal robustness analysis of network-wide railway delay prediction. Focusing on the Dutch railway network, this research utilizes interpretable tree-based ensembles to integrate granular topological, environmental, and operational features. The overarching finding establishes that while feature-rich tree-based models improve simultaneous (within-month) prediction, predictive performance systematically degrades when evaluated across non-simultaneous (future) months. Furthermore, multi-horizon SHAP and dispersion analyses explicitly link this degradation to environmental feature volatility and instability within the statistical target definition. Ultimately, this thesis demonstrates that richer feature sets alone are insufficient to resolve long-term forecasting constraints, underscoring the necessity to transition toward dynamic, season-aware architectures anchored by absolute operational boundaries.
View PDFOpen arXiv