BiLSTM-Based Train Delay Prediction: A Case Study of Major High-Speed Railway Stations in Shanghai

With the continuous expansion of high-speed railway networks and the increasing complexity of train operations, accurate and real-time train delay prediction has become essential for improving service quality and operational efficiency. This study proposes a short-term train delay prediction model based on Bidirectional Long Short-Term Memory (BiLSTM), focusing on five major high-speed railway stations in the Shanghai region. The model is trained and validated using real-world operational data, and its performance is evaluated with Mean Absolute Error (MAE) and loss metrics. Experimental results demonstrate that the proposed BiLSTM model outperforms benchmark models in prediction accuracy, convergence speed, and generalization capability, achieving a minimum validation MAE of 0.0929. Visualization of loss curves and prediction outputs further confirms the model’s superior learning ability and robustness. The proposed approach offers practical technical support for proactive delay management and intelligent scheduling in urban rail systems.