Analysis and Prediction of Factors Influencing Fatigue Driving in Freight Vehicles Based on Causal Analysis and GBDT Model

Fatigue driving of freight vehicles is a major threat to transport safety, often causing heavy casualties and property losses. However, existing studies only focus on superficial correlations between fatigue driving and influencing factors, failing to reveal intrinsic causal mechanisms, which limits practical guidance for prevention. To address this gap, this study, focusing on safety performance analysis in intelligent transportation systems and machine learning applications for sustainable transport management, uses monitoring data of “two types of passenger vehicles and one type of hazardous materials transport vehicle” in Shanghai. It identifies causal relationships between fatigue driving and 19 key factors (vehicle speed, driving time period, etc.) via a causal inference framework. Results show that 10 factors (including driving during specific periods) positively affect fatigue driving, while 9 factors (including vehicle speed) have negative effects. A Causal-GBDT Hybrid Model is built by weighting causal core factors into XGBoost (1.7.6) and CatBoost (1.2). Results show causal weights raise XGBoost accuracy from 90% to 93% and CatBoost from 89% to 94%. This clarifies fatigue triggers, provides technical support for targeted prevention, and advances machine learning in freight safety risk management. The research results can provide technical support for the development of real-time fatigue warning systems for freight vehicle and traffic safety management policies, contributing to the sustainable improvement of road transport safety.