A Hybrid LSTM-CPS Approach for Long-Term Prediction of Train Delays in Multivariate Time Series

In many big cities, train delays are among the most complained-about events by public. Although various models have been proposed for delay prediction, prior studies on both primary and secondary prediction limited in number. Recent advances deep learning approaches increasing availability of data sources has created new opportunities more efficient accurate prediction. this study, we propose a hybrid solution integrating long short-term memory (LSTM) Critical Point Search (CPS). LSTM deals with long-term tasks trains’ running time dwell time, while CPS uses predicted values nominal timetable to identify based causes, run-time delay, delay. To validate model analyse its performance, compare standard model. The results demonstrate that variants outperform LSTM, predicting steps feature. experiment also showed irregularities historical trends, which draws attention further research.