Improved Logit Model of Urban Rail Transit Passenger Flow Allocation Based on Mobile Signaling Data

In response to the increasing number of path choices in dense rail transit networks, a new calibration method is proposed to address allocation errors in the traditional logit model. A generalized cost function based on travel time is constructed, and actual passenger travel trajectories are matched with valid paths obtained from the K-shortest path algorithm. Utility factors, including travel time, walking time, and transfers, are considered. A calibrated logit model is used to determine the proportions of each path chosen, enabling accurate rail network passenger flow allocation. A case study in Shenzhen demonstrates high accuracy, with an average error of 3.5%, 91% of section errors below 10%, and 73% of cross-sectional errors less than 5%. This method provides a reference for traffic management in planning and operations.