Train Delay Predictions Using Markov Chains Based on Process Time Deviations and Elastic State Boundaries

Train delays are inconvenient for passengers and major problems in railway operations. When occur, it is vital to provide timely information regarding at their departing, interchanging, final stations. Furthermore, real-time traffic control requires on how propagate throughout the network. Among a multitude of applied models predict train delays, Markov chains have proven be stochastic benchmark approaches due simplicity, interpretability, solid performances. In this study, we introduce an advanced chain setting using historical operation data. Therefore, based process time deviations instead absolute relaxed commonly used stationarity assumptions transition probabilities terms direction, line, location. Additionally, defined state space elastically analyzed benefit increasing dimension. We show (via test case Swiss network) that our proposed model achieves prediction accuracy gain 56% mean error (MAE) compared state-of-the-art delays. also illustrate performance advantages training data sparsity.