The factors that cause traffic accidents are related to the characteristics of drivers. Based on K-means++, cluster analysis was carried out for the number of casualties, age, driving experience, accident responsibility, unsafe behavior value of drivers and unsafe state value of vehicles, so as to discuss the tendency of drivers with different characteristics to produce different types of traffic accident factors. The results show that when the number of clusters is 3, the silhouette coefficient is 0.52, and the drivers are divided into three groups. Among them: drivers with less driving experience, lower age and higher accident liability have higher unsafe behavior values. The group of drivers with higher driving experience, higher age and higher accident liability has a higher vehicle unsafe state value and a higher number of casualties. The unsafe behavior value and the unsafe state value of the vehicle are lower in the group of drivers with lower accident liability.