Spatiotemporal trip pattern analysis in a regional bus network with limited automated fare collection data: a case study of the Pioneer Valley

Understanding passenger trip patterns is crucial to enabling transit agencies plan effectively and equitably, particularly in response to disruptive events. However, current data collection systems either do not collect detailed passenger behaviour or are expensive to implement. In this paper, we exploit the relatively cheap big data collected by a mobile ticketing system to efficiently extract distinctive travel patterns. We employed our methods on a sample of anonymized data from the Pioneer Valley Transit Authority in Massachusetts. First, we applied a greedy approach to infer trip boarding stop locations. Then we computed the multi-dimensional dissimilarity of the passenger activation time series using the AWarp alignment algorithm, which works well with sparse data. Finally, we clustered these spatiotemporal patterns using hierarchical clustering. Our novel method has resulted in four spatiotemporal trip pattern typologies which we analysed based on demographics, hourly and daily distributions, faretypes, trip length and duration, among other metrics. Three of the typologies were associated with regular commuters, differentiated by either boarding time or transfer propensity. The fourth typology was mostly associated with leisure or other activities. Beyond yielding insights and facilitating demand estimation for planners in the study area, we expect that our framework can be readily applied to aid future decision-making efforts in similar study areas with minimal data availability.