Estimation of passenger route choices for urban rail transit system based on automatic fare collection mined data

When multiple routes connect the same origin and destination (OD), passenger route choices are related to the operation and management of urban rail transit, including the design of the train plan, fare clearing, and passenger distribution. The passengers' travel data from an automatic fare collection (AFC) system provides useful data for the analysis of passenger route choices. We propose an estimation method of passenger route choices based on AFC data, which includes passenger's travel time and OD. In this paper, the AFC data used for analysis were mined to eliminate unreasonable individual travel data. In this study, the components of passenger travel times and their influencing factors are analyzed and the results indicate that the travel times are related to the route choices of passengers. Cluster analysis indicated that passenger travel times follow the logarithmic normal distribution. Based on this and a multi-route distribution sample, an estimation model for passenger route choice proportions is proposed. Furthermore, experimental results show that the data mining method has high accuracy and the result also supported the hypothesis about the travel time distribution. The model was applied to the Chengdu metro, which is one of the largest metro systems in the world, and the obtained results indicate that the model works well. In addition to the initial application, the estimation method provides a new method to calculate the route choice proportion for urban rail transit with maximum dependability and timeliness.