Road rage recognition model for truck drivers based on K-means clustering and random forest algorithm

Road rage during driving have an essential impact on driving safety. With the aim of effectively identifying road rage in truck drivers, a driver anger scale (DAS) for truck drivers is developed based on the feature of truck drivers, and 400 valid questionnaires are collected from professional truck drivers. Next, by means of a K-means++ algorithm, truck drivers are divided into two types: high and low road rage. Finally, a road rage recognition model is established based on a random forest (RF) algorithm, the interpretable machine learning framework SHAP (SHapley Additive exPlanations) is introduced to extract the important influencing factors of road rage, and the features of road rage in truck drivers are further explored. The results show that the recognition accuracy of road rage based on the RF model is 76.3%, the recall rate is 92.5%, and the F1-score is 83.6%. Compared with decision tree (DT), support vector machine (SVM) and K-nearest neighbor (KNN), the recognition performance of the RF road rage recognition model is more reliable and stable. The important influencing variables on road rage are the driving environment, driving duration, driving experience, and violation frequency. The degree of road rage in truck drivers in a daytime driving environment is significantly higher than that at night. Driving experience is significantly negatively correlated with road rage, and truck drivers who are prone to fatigue driving and have a high frequency of traffic violations have a higher degree of road rage. These findings provide important insights into managing road rage among truck drivers, helping to develop effective strategies to reduce road rage and to enhance road transport safety.