Understanding passenger route choice behavior under the influence of detailed route information based on smart card data

Most previous studies explored the route choice behavior of metro passengers using stated preference (SP) survey data, but the SP data are inevitably subject to endogenous and selection bias. In contrast, automated fare collection (AFC) data record travel information for nearly all passengers at boarding and alighting stations. However, due to the seamless transfer in urban rail transit, it becomes challenging to track the actual routes of passengers accurately using AFC data. Fortunately, based on a data-driven method, the chosen route and detailed travel information (e.g., segmented travel time, train load status) can be inferred with AFC data. To fill the research gaps, this paper delves into the route choice mechanism by considering the effect of detailed route information, taking Nanjing Metro, China as a case study. A Conditional Multinomial Logit model is employed to examine the effect of determinants on route choice behavior for metro passengers. The results show that the route choice model considering dynamic segmented travel time and train load status has better fit performance than the benchmark models. The sensitivity of the walking time is found to be similar to that of in-vehicle time for metro passengers, but a stronger distaste for waiting time or queuing time is observed. Besides, the crowding-related attributes are negative for route choice, but Nanjing Metro passengers present a higher tolerance for crowding compared with passengers in developed countries. These findings provide an accurate and comprehensive insight into the route choice behavior of metro passengers.