Multi-Metric Weighted Ensemble Learning for Train Delay Prediction

Golchehreh, Masoomeh, Parr, Gerard ORCID: https://orcid.org/0000-0002-9365-9132 and Wang, Wenjia (2026) Multi-Metric Weighted Ensemble Learning for Train Delay Prediction. In: The 11th int. conference on Machine Learning Technology, 2026-05-21 - 2026-05-23.

Preview

PDF (Multi-Metric Weighted Ensemble Learning for Train DelayPrediction) - Accepted Version Available under License Creative Commons Attribution. Download (2MB) | Preview

Abstract

Train delays are a common problem in railway operations. Accurate delay prediction with machine learning can help train operators to adjust their operational decisions in time to prevent or reduce delays. % predictive performance often varies across routes, sections, and operating conditions. This paper proposes a Multi-Metric Weighted Ensemble Method (MWEM), a heterogeneous ensemble that integrates diverse learning models through adaptive and performance-driven weighting for station-to-station delay prediction. The key idea is a metric-informed cross-validation scheme that assigns ensemble weights from validation behaviour under multiple criteria to avoid fixed or heuristic configurations and allow the combination of complementary model strengths. The MWEM is evaluated on two real-world UK railway routes and compared against individual baseline models and existing state-of-the-art ensembles. The results show %consistent gains across network sections, that the MWEM achieved the best and consistent overall accuracies, significantly better than all the other models. % relative to the strongest single model (reducing RMSE by 8\%). These findings demonstrate that multi-metric, data-driven weighting provides an effective approach for improving the reliability of delay prediction in practical railway settings.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	machine learning,artificial intelligence,train delay prediction,artificial intelligence ,/dk/atira/pure/subjectarea/asjc/1700/1702
Faculty \ School:	Faculty of Science > School of Computing Sciences
UEA Research Groups:	Faculty of Science > Research Groups > Data Science and AI Faculty of Science > Research Groups > Health Computing Faculty of Science > Research Groups > Cyber Intelligence and Networks
Depositing User:	LivePure Connector
Date Deposited:	21 Oct 2026 09:34
Last Modified:	21 Oct 2026 09:34
URI:	https://ueaeprints.uea.ac.uk/id/eprint/104825
DOI:

Downloads

Actions (login required)

View Item