Identifying spatio-temporal patterns of bus bunching in urban networks

Abstract The objective of this paper is to identify hot spots of bus bunching events at the network level, both in time and space, using Automatic Vehicle Location (AVL) data from the Athens (Greece) Public Transportation System. A two-step spatio-temporal clustering analysis is employed for identifying localized hot spots in space and time and for refining detected hot spots, based on the nature of bus bunching events. First, the Spatio-Temporal Density Based Scanning Algorithm with Noise (ST-DBSCAN) is applied to distinguish bunching patterns at the network level and subsequently a k++means algorithm is employed to distinguish different types of bunching clusters. Results offer insights on specific time periods and route segments, where bus bunching events are more likely to occur and, also, on how bus bunching clusters change over time. Further, headway deviation analysis reveals the differences in the characteristics of the various bunching event types per line, showing that routes running on shared corridors experience more issues while underlying causes may vary per line. Collectively, results can help guide practice toward more flexible solutions and control strategies. Indeed, depending on the type of spatio-temporal patterns detected, appropriate improvements in service planning and real-time control strategies may be identified in order to mitigate their negative effects and improve quality of service. In light of emerging electric public transport systems, the proposed framework can be also used to determine preventive strategies and improve reliability in affected stops prior to the deployment of charging infrastructure.