Research on lane-changing decision model with driving style based on XGBoost

Correct driving decisions are a prerequisite for safe driving. In this paper, we construct a lane-changing decision model considering driving style in order to provide drivers with more accurate and timely lane-changing decision interventions. Based on NGSIM data, two-step trajectory reconstruction technique is firstly used to deal with data outliers and noise, and then vehicle lane-changing trajectory extraction and phase classification are carried out according to the definition of lane-changing decision, and k-means clustering method is also used to realize driving style calibration and determine lane-changing decision variables based on following phase data. The final XGBoost lane-changing decision model was constructed, and the need to incorporate driving style features and the superiority of XGBoost over RF and SVM models were verified through comparative experiments.