TAXI REFUELING BEHAVIOR ANALYSIS BY USING TRAJECTORY DATA BASED ON MACHINE LEARNING

With the continuous development of GPS technology and internet of vehicles applications, the taxis have been equipped with mobile terminals-GPS devices in many cities, making a large-scale acquisition of vehicle trajectory data possible. The data itself does not have obvious temporal and spatial characteristics. Taxi refueling behavior can reflect the energy needs of urban public transport, and whether the relation between the layout of urban public resources and the demand is reasonable. In order to study the characteristics of the taxi's refueling behavior, this paper makes use of the GPS trajectory data of taxis in Wuhan, and is based on Decision Tree (DT) method training the classification model, which is evaluated and analyzed. The refueling behavior of Wuhan is extracted to analyze the temporal and spatial distribution characteristics of the behavior.