Cause Analysis of Bus Driver Fatigue Driving Level with DMS Warning-Based Data

Previous research indicated a strong correlation between the most traffic crashes and drivers’ fatigue driving behavior. In this paper, we investigated the influence of driver, vehicle, road, and environmental factors on bus drivers’ fatigue driving level using machine learning models and the SHAP interpretable machine learning framework. The study integrated data from driver monitoring system (DMS) warnings and GPS, along with information on driver attributes, weather conditions, and road characteristics. A comparative analysis was conducted among XGBoost and Random Forest. The accuracy demonstrated the outstanding performance of XGBoost in identifying fatigue driving level. The top five influential factors were identified as speed, acceleration, period, age, and location. This research provides valuable technological support for management of driver fatigue and for improving bus safety operations.