Ridership Prediction of Urban Rail Transit Stations Based on AFC and POI Data

Abstract

Ridership prediction of urban rail transit stations is of great significance for the operation and management of rail transit and configuration of facilities around stations. This study used automatic fare collection (AFC) data of the rail transit in Nanjing, China, for a month to obtain station ridership. Based on the point of interest (POI) data (within 800 m around urban rail transit stations), built environment factors such as land type and station accessibility were extracted, and a variable set of built environment factors was then established. Multiple collinearity and spatial autocorrelation analyses were used to screen the variables used in the regression model. A geographically weighted regression (GWR) model was constructed to explore the spatial heterogeneity of the influence on ridership of the built environment around the urban rail stations and to predict ridership. The results show that the GWR model can effectively capture the spatial heterogeneity of the effect of built environment factors on station ridership, and its ridership prediction accuracy is significantly better than that of the ordinary least squares model.