Transportation Risk Assessment for Hazardous Chemical Road Transport Companies Using an Extreme Gradient Boosting Model

Abstract

Road accidents involving vehicles carrying hazardous chemicals can result in significant loss of life and property. Risk factors that relate specifically to road transport companies who handle hazardous chemicals are closely associated with such road accidents. Little research has been done to analyze the transport process, assess the transport risk, or suggest bespoke mitigation and management measures for hazardous chemical road transport companies. To this end, this study combined the original Extreme Gradient Boosting model with Bayesian optimization (XGBoost-BO) to assess and mitigate transport risk in individual hazardous chemical transport companies. The original Extreme Gradient Boosting model was combined with Bayesian optimization to assess and mitigate transport risk in hazardous chemical transport companies. Twelve indicators related to driver behavior, transport operations, and onboard equipment data quality formed the metrics framework. The Bayesian optimization tuned the various parameters in the Extreme Gradient Boosting model. Shapley additive explanations (SHAP) values were introduced to visualize and interpret the model results. The proposed model demonstrates superior performance over other machine learning models in all measures of effectiveness. The major factors affecting risk were the mileage of vehicles, the number of waybills, and the duration of fatigued driving. The proposed model can provide bespoke management recommendations for hazardous chemical transport companies—companies can learn from their values of indicators and take apposite measures to mitigate their unique risk factors—thus enhancing overall transportation safety.