Analyzing and Predicting Railway Operational Accidents Based on Fishbone Diagram and Bayesian Networks

Abstract

The prevention of railway operational accidents has become one of the leading issues in railway safety. Identifying the impact factors which significantly affect railway operating is critical for decreasing the occurrence of railway accidents. In this study, 8440 samples of accident data are selected as the datasets for analyzing. Fishbone diagram is applied to obtain the factors which cause the accident from the perspective of human-equipment-environment-management system theory. Then, the Bayesian network method was selected to establish a railway operation safety accident prediction model, and the sensitivity analysis method was used to obtain the sensitivity of each variable factor to the accident level. The results show that season, location, trouble maker and job function have a significant impact on railway safety, and their sensitivity was 0.4577, 0.4116, 0.3478 and 0.3192, respectively. Research helps the railway sector to understand the fundamental causes of accidents, and provides an effective reference for accident prevention, which is conducive to the long-term development of railway transportation.