PREDICTION OF REAL-TIME TRAIN ARRIVAL TIMES ALONG THE SWEDISH SOUTHERN MAINLINE

Abstract

Real-time train arrival time prediction is crucial for providing passenger information and timely decision support. The paper develops methods to simultaneously predict train arrival times at downstream stations, including direct multiple output liner regression (DMOLR) and seemingly unrelated regression (SUR) models. To capture correlations of prediction equations, two bias correction terms are tested: (1) one-step prior prediction error and (2) upstream prediction errors. The models are validated on high-speed trains operation data along the Swedish Southern Mainline from 2016 to 2020. The results show that the DMOLR model slightly outperforms the SUR. The DMOLR’s prediction performance improves up to 0.32% and 24.03% in term of RMSE and R 2 respectively when upstream prediction errors are considered.