Comprehensive Identification of Taxi Driving State Based on Kinematic Feature Pattern

Abstract:

Based on GNSS data， the rule and method of extracting taxi kinematic segments were proposed. According to the principal component analysis （PCA） and cumulative contribution rate， 8 key indicators were determined to represent kinematic segments. Combined with K-mean clustering algorithm， the taxi kinematic feature patterns were obtained. In order to ensure the objective rationality of the weight of key indicators， the CRITIC method considering the correlation of indicators and the entropy weight method considering the dispersion degree of indicators were adopted. The VIKOR evaluation model based on Nash equilibrium combination weight was established to evaluate the driving state of the taxi under multi spatio-temporal scenarios. The results show that the combined weighting method based on the Nash equilibrium can effectively integrate the merits of the CIRTIC method and entropy weight method and obtain a more reasonable weight coefficient. In terms of overall safety， efficiency and comfort， the driving condition of the main road and secondary road is better than that of the branch road. The safety of taxi driving is the best in the morning peak， and relatively ordinary in the evening peak and flat peak.