Evaluation of Safe Driving Behavior of Transport Vehicles Based on K-SVM-XGBoost

In recent years, the total number of transport vehicles in China is increasing by a large margin, strengthening traffic safety management to effectively prevent accidents has become the focus of traffic research. The driver as the operator of vehicle, its driving behavior is the most important factor affecting the traffic safety. Therefore, it is of great significance to study driving behavior to prevent traffic accidents and promote the development of intelligent traffic and driving assistance system. Based on the data of the Internet of vehicles, this paper builds a SVM-based driving behavior health evaluation model, which extracts the feature values to obtain the optimal classification hyperplane, and then uses the geometric interval as the evaluation index of driving behavior health. At the same time, taking driving behavior, road, efficiency, energy saving and weather factors with different weights into consideration, it divides driving behavior into four levels: excellent, average, qualified and unqualified level, based on KMeans and SVM, and then XGBoost training model is used. The experiment concluded that the average accuracy is 99.21% and the average recall rate is 98.5%. These results proved that the model is effective and feasible.