Abstract: Existing causes analysis of coal mine accidents is mainly based on unilateral factor, ignoring correlation and integrity of four factors of human, machine, environment and management, and the types of accidents involved are relatively generalized. In view of the above problems, factors inducing coal mine gas explosion accident are selected from four aspects of human, machine, environment and management, and correlation analysis is used to select variables with strong correlation. Bayesian network model of coal mine gas explosion causes is constructed by GeNie, and reliability and accuracy of the model are verified by cross-validation method. The parameters of each node in the model are analyzed by Bayesian network parameter learning and sensitivity analysis. The conditional probability distribution and posterior of the relevant nodes under different conditions are calculated to extract key factors inducing gas explosion accidents in coal mines. The analysis results show that insufficient ventilation will greatly increase the possibility of excessive gas content in coal seams, unqualified employee training is the main cause of gas missed inspection; in the case of coal mine gas explosion accidents has happened, the most likely cause is excessive gas content, and followed by missed gas inspection; the most critical factors leading to coal mine gas explosion accidents are excessive gas content, missed gas inspection, unstable roof, and unsound laws and regulations.