Calibrating travel time thresholds with cluster analysis and AFC data for passenger reasonable route generation on an urban rail transit network

Estimating the route choice patterns for transit passengers is important to improve service reliability. The size and composition of a route choice set affects the choice model estimation and passenger flow calculations for urban rail transit (URT) networks. With the existing threshold decision method, there will be omissions or excess routes in the generated route set, which lead to a significant deviation in passenger flow assignments. This paper proposes a data-driven approach to calibrate the travel time thresholds when generating reasonable route choice sets. First, an automatic fare collection (AFC) data-driven framework is established to more accurately calibrate and dynamically update travel time thresholds with changes in the URT system. The framework consists of four steps: data preprocessing, origin–destination-based threshold calculation, cluster analysis-based calibration, and calibrated result output and update. Second, the proposed approach is applied to the Beijing subway as a case study, and several promising results are analyzed that allow the optimization of existing travel time thresholds. The obtained results help in the estimation of route choice behavior to validate current rail transit assignment models. This study is also applicable for other rail transit networks with AFC systems to record passenger passage times at both entry and exit gates.