Research on Qualitative Classification Method of Drivers' Driving Style

Driving style shows driving behavior in daily life of driver. The study of driving style can predict the driving behavior in the real environment to improve the accuracy of driver intention recognition and decision model, and it can guide the improvement of driving habits, as well as improve driving safety and fuel economy. Qualitative analysis is an important part of driving style classification. At present, most studies are based on a single mature scale for verification and simplification to meet the requirements. The main problems are that the assessment dimension is not accurate, the sample validity cannot be guaranteed, and the driving style cannot be reflected comprehensively. In this paper, the three dimensions of conservative, calm and radical were determined according to the requirements of the project. The self-made scale was established by referring to the MDSI, DBQ, DSAS and GDMS. The validity of samples were ensured by cross validation and the self-made scale was revised based on the reliability and validity verification. Then, an effective driving style classification scale related to the assessment dimension was formed. The appropriate clustering method was determined by comparing the contour coefficients corresponding to the classification results of the three methods of K-means clustering, hierarchical clustering and PCA-based dimensionality reduction clustering. Finally, a qualitative classification method for effective driving styles was formed for the three dimensions of conservative, calm and radical.