Pick-Up Time Analysis and Prediction for Carsharing Users Based on Decision Tree

The development of carsharing can reduce the number of private cars, which can save resources. Due to the limited supply of vehicles and diversified demands of users, it is necessary to plan the temporal and spatial distribution of cars. Predicting the pick-up time of carsharing users is of great significance to understand the travel preference of carsharing users, which can help operators formulate operational strategies such as relocation and pricing. To this end, this study adopts an improved decision tree (DT) to analyze and predict pick-up time for carsharing users. Firstly, the ordered clustering method is used to discretize time. Secondly, the random forest (RF) model is constructed to extract key features. Finally, the model of the C5.0 DT is constructed to predict the pick-up time of users. A case study is conducted to demonstrate the proposed model. The results indicate that the prediction accuracy of users’ pick-up time can reach 87%. The characteristic of pick-up time of carsharing users is clearly analyzed.