Abstract: The utilization rate of metro accident data in China is insufficient, which makes it difficult to provide theoretical support and help for the prevention before accidents, site decision making at the time of accidents, analysis and responsibility determination after accidents. Based on HFACS-MA, the accident cause analysis model is divided into accident cause layer and accident feedback layer, and factor analysis on 30 metro accident cases are carried out. Apriori algorithm is adopted to analyze the correlation degree of human causes to accidents, and accident cause chain diagram is drawn. Results show that the organization influence layer in the accident cause chain involves the most factors; weak supervision is the main connecting factor that needs to be paid attention to; and untimely handling of accidents is the most direct cause leading to the expansion and spread of accidents.