Abstract: In order to conduct in-depth research on the factors influencing the severity of truck accidents on mountainous freeways and achieve active and precise prevention and control of traffic accidents, this paper selects collision type features, vehicle type features, road features, and environmental features as input variables, and accident severity as binary output variables. Three machine learning models, including decision tree model (DT), random forest model (RF), and support vector machine model (SVM), are constructed. Evaluate the quality of the model based on accuracy, precision, recall, and F1 indicators, and use SHAP method to deeply analyze the output results of the machine learning model. The research results indicate that the RF model is superior to the DT model and SVM model. From the perspective of influencing factors, the variables of overturning, no slope, cement road surface, turning, frontal collision, accident time from 19:00 to 6:59, and no roadside protective measures have a significant impact on the severity of truck accidents on mountainous freeways.

Key words: traffic safety, mountainous freeways, accident severity, truck accidents, machine learning