Abstract: At present, road transportation accidents of hazardous chemicals occur frequently in China. To reduce such accidents and enhance system safety, this paper focuses on accident prediction. In view of accident prediction, it adopts a BP neural network optimized by the improved particle swarm optimization (PSO) algorithm based on collected accident data to predict the severity of hazardous chemicals road transportation accidents in China and analyze the accident causes. The results indicate that severe accidents are more likely to occur during the early morning hours, under poor visibility or adverse weather conditions such as fog, and when transporting hazardous chemicals such as explosives and flammable liquids in road transportation. Case validation demonstrates that the PSO-BP neural network model achieves an average prediction accuracy of 84.9% and it can effectively correct errors of the single model. This study can provide more reliable data support for analyzing and assessing safety trends in road transportation of hazardous chemicals, and offer assistance in the formulation of accident prevention and control strategies as well as decision-making.

Key words: road transportation of hazardous chemicals, accident prediction, BP neural network, particle swarm optimization