Driving Style Analysis for Car-sharing Drivers with Low-frequency Trajectory Data

Abstract:

Driving style analysis was conducted based on the operation data of a car-sharing project located in Shanghai. Rather than high resolution driving behavior data in most driving style studies, low-frequency trajectory data were utilized. The relative speeding time ratio and its coefficient of variation on urban expressways were used as feature variables. K-means clustering algorithm was used to classify driving styles. A total of three categories were concluded, which are calm, moderate and aggressive with the percentages of 54.04%, 36.60%, and 9.36% correspondingly. Then, for the purpose of understanding different driving styles, comparison analyses were further conducted from the aspects of trip characteristics, vehicle operation features, and personal information. The results show that drivers with distinct styles have substantial differences in their trip and vehicle operation characteristics. The aggressive drivers tend to drive faster, holding higher speeding tendency but better vehicle energy efficiency. Besides, no statistically significant differences in age, gender or violation between driving styles are identified.