Decomposing Metro-Bus Transfer Time with Smart Card Data

Metro-bus transfer is a crucial step in last mile urban mobility and is where riders can waste significant amount of time. While total transfer time is a good measure for the integration of metro and bus services, it cannot tell how much time is spent for transferring and waiting, respectively. Therefore, a deeper understanding of transfer behavior and service qualities requires decomposing this metro-bus transfer into two legs: (1) walking from metro to bus stop and (2) waiting at bus stop. This study proposes to model total transfer time with exponentially modified Wald (ex-Wald) distribution, which estimates walking time via an inverse Gaussian distribution and waiting time via exponential distribution, respectively. Experimental results on a real-world smart card data set show that the proposed distribution describes the data more robust than other baseline models.