Real-time High-Speed Train Delay Prediction using Seemingly Unrelated Regression Models

Understanding the impact of various factors on train arrival delays is a prerequisite for effective railway traffic operating control and management. Existing studies model the train delay prediction problems using a single, generic equation, restricting their capability in accounting for heterogeneous impacts of spatiotemporal factors on arrival delays as the train travels along its route. The paper proposes a set of equations conditional on the train location for predicting train arrival delays at stations. We develop a seemingly unrelated regression equation (SURE) model to estimate the coefficients simultaneously while considering potential correlations between prediction residuals caused by shared unobserved variables (e.g., driver characteristics). The operational data for high-speed trains on Sweden's Southern Mainline from 2016 to 2020 is used to validate the proposed model and explore the effects of operation-related factors on train arrival delays. The results confirm the necessity of developing a set of station-specific delay prediction models to understand the heterogeneous impact of explanatory variables, and SURE provides more efficient parameter estimations than the traditional ordinary least squares regression (OLS). The important factors impacting train arrival delays include the scheduled and actual running time, scheduled dwell time, and train arrival delays at preceding stations. However, the impact of these factors could vary depending on where the station is, and different types of operating management strategies should be targeted.