Stop-Level Bus Clustering Prediction Incorporating Cross-Route Operational Context in Shared Corridors

Bus clustering, where consecutive buses arrive at stops in rapid succession instead of maintaining planned headways, is a major source of unreliability in public transport systems. This paper proposes a stop-level machine learning framework for predicting bus clustering events using fully open operational and environmental data. The approach integrates real-time vehicle location data and scheduled timetables from the UK Bus Open Data Service with weather data. Using a major inter-urban route in West Yorkshire (UK) as a case study, clustering is predicted at individual stops across the full corridor rather than at aggregate route level. Further, the work examines the importance of cross-route operational context from neighbouring services. The best-performing model achieves an F1 score of 0.90 and ROC-AUC of 0.995, while cross-route context provides modest improvements.