A Network-Based Model of Passenger Transfer Flow between Bus and Metro: An Application to the Public Transport System of Beijing

In a multimodal public transport network, transfers are inevitable. Planning and managing an efficient transfer connection is thus important and requires an understanding of the factors that influence those transfers. Existing studies on predicting passenger transfer flows have mainly used transit assignment models based on route choice, which need extensive computation and underlying behavioral assumptions. Inspired by studies that use network properties to estimate public transport (PT) demand, this paper proposes to use the network properties of a multimodal PT system to explain transfer flows. A statistical model is estimated to identify the relationship between transfer flow and the network properties in a joint bus and metro network. Apart from transfer time, the number of stops, and bus lines, the most important network property we propose in this study is transfer accessibility. Transfer accessibility is a newly defined indicator for the geographic factors contributing to the possibility of transferring at a station, given its position in a multimodal PT network, based on an adapted gravity-based measure. It assumes that transfer accessibility at each station is proportional to the number of reachable points of interest within the network and dependent on a cost function describing the effect of distance. The R-squared of the regression model we propose is 0.69, based on the smart card data, PT network data, and Points of Interest (POIs) data from the city of Beijing, China. This suggests that the model could offer some decision support for PT planners especially when complex network assignment models are too computationally intensive to calibrate and use.