Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{178475,
        author = {Pushpalatha.A and Arish A and Chinrasu S and Manojkumar S and Muthuraman P},
        title = {Train Delay Prediction Using Machine Learning},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {11},
        number = {12},
        pages = {5040-5046},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=178475},
        abstract = {Train transport systems are integral parts of urban mobility and national logistics; However, unexpected delay often disrupts operations, reduces efficiency, and affects passenger satisfaction. Traditional approaches to delay the management struggle for the dynamic and interaction nature of the railway network. This paper introduces a three-pronged prediction framework designed to enhance the reliability and accuracy of train delay forecasts. A hybrid sequence-regression model is developed, where Long Short-Term Memory Extreme Gradient Boosting (LSTM+ XGBoost) networks analyse the data to extract the output, and  XGBoost refines predictions based on contextual framework such as weather conditions, train type, and station congestion. To find the spatial delay propagation, a Graph Neural Network (GNN) models the railway network as a graph, enabling the system to forecast delay cascades across interconnected stations. Additionally, a Transformer-Based Attention (TBA) mechanism is employed to improve the interpretability of the predictions by prioritizing significant temporal features and external influences. Simulated and real-world datasets from public railway systems are used for training and evaluation. Results achieve 91% classification accuracy and reduce the Mean Absolute Error (MAE) by 18% compared to conventional methods. The findings confirm that integrating temporal modelling, spatial reasoning, and attention-based mechanisms significantly improves delay prediction performance in complex rail environments.},
        keywords = {Train Delay Prediction, LSTM, XGBoost, GNN, Transformer Attention, Time Series Forecasting, Railway Scheduling, Machine Learning, Transport Analytics.},
        month = {May},
        }