Improving Bus Trip Generation Modeling Using Mobile Payment Data: An Empirical Study

Abstract

Smart card data (SCD) have been widely used to estimate the effect of built environment on transit ridership. However, most studies have ignored the remaining parts of trips paid in cash, whose share is also sizable, and gives rise to estimation bias in bus travel demand generation. Luckily, the availability of mobile payment replaces most cash-based trips, which renders us to track these trip records. Therefore, this study develops a series of models considering overall, spatial, temporal, and nonlinear effects of built environment on bus ridership. An empirical study is conducted in Nanjing, China, from four weeks of SCD and mobile payment data (MPD) in the bus transit system. The results suggest that mobile payment users generally use bus services less frequently per week compared to smart card users. The inclusion of MPD reveals significant spatial variations in ridership, particularly across accessibility variables. Temporal analysis shows that incorporating MPD is most beneficial in reducing estimation bias during morning and evening peak hours. Furthermore, nonlinear models demonstrate more pronounced effects when both SCD and MPD are considered, especially in identifying threshold effects of the built environment on ridership. This study underscores the importance of adopting new payment methods in transit demand modeling to enhance the accuracy and robustness of bus ridership predictions, providing deeper insights into urban mobility patterns.