The causal effect of unsafe driving behavior of heavy-duty trucks on accident casualties

This paper aims to investigate the causal effects of unsafe driving behaviors on the casualty rates of heavy-duty truck transportation accidents and examines the heterogeneity of accident influencing factors. Utilizing correlation analysis and Recursive Feature Elimination (RFE), we screened 17 confusion variables from the 2016-2021 heavy-duty truck transportation accident data in the US Crash Report Sampling System database. In this study, LightGBM, XGBoost, Random Forest (RF), and Support Vector Machine (SVM) models were employed as estimators to construct a causal inference model of the impact of unsafe driving behavior on accident casualties, based on Double Robust Learning. The findings reveal that the LightGBM model exhibits the best overall performance, with a prediction accuracy ranging from 0.79 to 0.90; Drivers of heavy-duty trucks who fail to comply with traffic signs and signals, engage in speeding, tailgating, and distracted driving, respectively elevate the risk of accidents and injuries by 21.75%, 18.15%, 17.08%, and 15.12%; Drivers under the age of 20 exhibit a higher risk of injury accidents as a result of unsafe driving behaviors, compared to other age groups; Similarly, driving on curved sections, inclines, and two-way non-segregated left-turn lanes is associated with an elevated risk of injury accidents due to unsafe driving. These findings can serve as a reference for the development of safety management strategies for heavy-duty trucks and for the reduction of casualty rates in truck-related accidents.