Using Point of Interest Data from Electronic Map to Predict Transit Station Ridership

Land use is the original source of all travel, and land use characteristics play a dominant role in direct ridership models. Among all land use characteristics, density and diversity are believed to be the most effective ones. Nevertheless, nearly all existing models describe land use patterns and distribution with two rough indexes, population and employment, while describing mixed land use with entropy index, diversity index, jobs-housing index, and so forth. These indexes fail to describe the details of different types of land use, and they are not capable of revealing the relationship between transit ridership and land use characteristics in a transit station catchment area. In order to solve this problem, POIs (points of interest) extracted from a Beijing electric map were chosen to describe land use characteristics. Combined with the delineation of transit station catchment areas, the precise land use characteristics of each catchment area were described with a corresponding POI built area. Based on the precise land use characteristics and station ridership data, a direct transit ridership model considering land use density and diversity was built using multiple regression analysis. The moderately high correlation coefficient, R2 14 =0.735, proves the remarkable potential of using POIs in transit ridership forecasting. In order to prove the feasibility of the direct model proposed in this paper, a model validation was carried out; validation results meet the initial expectations, proving that the direct ridership model developed in this paper could reflect this correlation between precise land use characteristics and station ridership.