FI-TW: An Open Train-Weather Dataset for Railway Delay Analysis in Finland. Vinicius Pozzobon Borin, Jean Michel de Souza Sant’Ana, Usama Raheel, Nurul Huda Mahmood. arXiv preprint, cs.LG, CC BY 4.0, January 2026, revised August 2026. DOI 10.48550/arXiv.2601.16592.
I co-authored this preprint, which introduces the first publicly available dataset linking Finnish railway operations with synchronized meteorological data, 2018 to 2024: approximately 38.5 million observations across the 5,915 km rail network and 209 weather stations.
Alongside the dataset, we established a baseline: an XGBoost model reaching 2.73-minute mean absolute error for station-level delay prediction, demonstrating the dataset’s value for machine-learning applications in railway operations.
The work grew out of my research internship at the Centre for Wireless Communications, University of Oulu, on the project “Artificial Intelligence for Predictive Modelling of Weather-Induced Delays in Finland’s Railway System”, supervised by Vinicius Pozzobon Borin, PhD, and Dr. Nurul Huda Mahmood.
A note on the title: the preprint was retitled on revision. The January 2026 v1 appeared as Integrating Meteorological and Operational Data: A Novel Approach to Understanding Railway Delays in Finland; the August 2026 v2 carries the FI-TW title above.