Understanding citywide transit patterns is important for transportation management, including city planning and route optimization. The wide deployment of automated fare collection (AFC) systems in public transit vehicles has enabled us to collect massive amounts of transit records, which capture passengers’ traveling activities. Based on such transit records, origin–destination associations have been studied extensively in the literature. However, the identification of transit patterns that establish the origin–transfer–destination (OTD) associations, in spite of its importance, is underdeveloped. In this paper, we propose a framework based on transit tensor factorization (TTF) to identify citywide travel patterns. In particular, we create a transit tensor, which summarizes the citywide OTD information of all passenger trips captured in the AFC records. The TTF framework imposes spatial regularization in the formulation to group nearby stations into meaningful regions and uses multitask learning to identify traffic flows among these regions at different times of the day and days of the week. Evaluated with large-scale, real-world data, our results show that the proposed TTF framework can effectively identify meaningful citywide transit patterns.
The online supplement is available at
https://doi.org/10.1287/ijoc.2018.0824
....

Transit Pattern Detection Using Tensor Factorization

10.1287/ijoc.2018.0824