Exploring Microscopic Driving Behaviors Using Large-Scale GPS Trajectory Data

The integration of large-scale trajectory data sets to conduct in-depth research on driving behavior is critical to traffic safety management. Based on large-scale GPS trajectory data, we obtain three time periods and extract four characteristic parameters of driving behavior and perform Principal Component Analysis (PCA). Based on Gaussian Mixed Model (GMM), two principal components are used as clustering basis. Then, the driving risk levels of cautious and aggressive drivers are defined, and the changes and differences in driving behavior of different drivers are analyzed. We further compare the proportion of different driver types and driving behaviors on weekdays and weekends and under different weather conditions. In addition, through the visualized spatial-temporal data distribution map, the spatial distribution changes of different driver types are analyzed. Results of this study are valuable for safety analysis of the individual driving behavior.