Abstract: In order to effectively reduce the occurrence of coal mine safety accidents, and to formulate scientific measures of disaster prevention, this study takes the relevant data of coal mine safety accidents in recent 11 years as statistical analysis samples, and studies the rules and characteristics of coal mine safety accidents in China by making analysis of two elements which are accident grades and types. With the occurrence of gas accidents, blasting, water hazards, transportation, roof, electromechanical, fire and other accidents as sample data, the grey neural network online prediction model was constructed and verified based on the data of 2021. The results show that general accidents are the most frequent, followed by larger accidents and major accidents; the number of roof, transportation, electromechanical and other accidents shows an overall upward trend, and the roof accidents are the most frequent; the mean relative error and root mean square error of the grey neural network model are 0.161 and 2.902, respectively, which are reduced by 0.234 and 2.945 compared to the grey model.Therefore, the grey neural network model is used to predict coal mine safety accidents, with higher accuracy and better stability.