Abstract: To achieve a quantitative assessment of the safety efficiency of heavy truck drivers, this paper proposes a framework for safety efficiency assessment of heavy truck drivers based on data envelopment analysis, which aims to identify key risk behaviors and provide optimization suggestions. First, the aggressive behavior and abnormal state data of drivers were obtained based on on-board devices, and a risk indicators database was constructed. Second, the safety efficiency evaluation of drivers was realized through data envelopment analysis, effective drivers and inefficient drivers in the group were determined, and the slack variables of the model were used to determine the main improvement direction of each inefficient driver. Finally, the benchmark decision making unit of drivers was constructed through risk variables and reference sets, which provided quantifiable suggestions for optimizing the risk behavior of inefficient drivers. The results can effectively distinguish the safety efficiency of heavy goods vehicle drivers and reveal the risk behaviors of inefficient drivers that need to be improved, so as to improve the driving operation of drivers. The results can effectively supervise the driving performance of drivers and achieve targeted and differentiated intervention training according to individual characteristics.