Data-Driven Real-Time Denied Boarding Prediction in Urban Railway Systems

Providing real-time crowding information in urban railways would enable informed travel decisions and encourage cooperative behavior of passengers, as well improve operating efficiency safety. However, the problem prediction is not trivial because unavailability ground-truth data, particularly for direct impact on passengers (e.g., denied boarding platforms). This paper proposes a data-driven method railway systems using automated fare collection (AFC) vehicle location (AVL) data. It predicts probability distribution or its derived metrics function explanatory variables, including demand, operations, incident-related factors. The validated through case study covering 18 months Hong Kong Mass Transit Railways. results highlight model’s accurate robust performance predicting platforms purely AFC AVL data an average 6%–7% error) under both recurrent non-recurrent situations, which transfer demand-related factors contribute most to prediction. model outputs can support proactive operations control management, provision customer without relying expensive time-consuming efforts. Customer information, example, include number trains wait before waiting time platform.