A novel Markov model for near-term railway delay prediction

Predicting the near-future delay with accuracy for trains is momentous railway operations and passengers’ traveling experience. This work aims to design prediction models train delays based on Netherlands Railway data. We first develop a chi-square test show that evolution over stations follows first-order Markov chain. then propose model non-homogeneous chains. To deal sparsity of transition matrices chains, we novel matrix recovery approach relies Gaussian kernel density estimation. Our numerical tests this outperforms other benchmark approaches in accuracy. The chain also shows be better than widely-used time series respect both interpretability Moreover, our proposed does not require complicated training process, which capable handling large-scale forecasting problems.