Due to its intelligence and convenience, the online car-hailing has developed rapidly. With the help of onboard GPS devices, GPS data will be collected during the operation cars. In this paper, we use the method of nuclear density estimation and K-Means clustering to analyze Xian GPS data of the week of October 2016. The distribution characteristics of Didi GPS data in time and space are analyzed from the perspective of drivers and passengers. The study indicates that on Friday, Saturday and Sunday, total orders per day and average orders per driver are higher than other days. The hotspots are similar within a week. It forms a hotspot distribution pattern of three residential areas plus two commercial areas. The study reveals the travel rules of Didi passengers in Xi'an, providing a reference to guiding Didi drivers to travel through areas with high demand and reducing empty cars.

CICTP