Abstract: Coal is the main energy source in China. It plays a crucial role in promoting economic and social development, and is the guarantee of China's energy security. This paper establishes a prediction model for the number of deaths in coal mine accidents under the influence of multiple factors through data analysis and modeling. Also, it predicts the safety production situation under the optimization of coal production capacity. Based on the statistical data of coal mine production capacity and the coal mine accident fatalities in China since 1949, the relationship between production capacity and accidents is analyzed. The regional characteristics of coal mine safety production situation is studied and a Long Short-Term Memory Network (LSTM) model is developed for predicting the number of deaths in coal mine accidents. And based on this model, the number of deaths in coal mine accidents under the influence of regional production capacity optimization policies was predicted. The research results show that ① based on China's coal production, accident fatalities, and million-ton mortality rate over the years, the development of the China's coal industry can be divided into five stages: initial establishment stage, exploration stage, recovery and rectification stage, institutional mechanism reform stage, and safe and efficient development stage. The national coal mine safety situation varies greatly in different periods due to the changes in safety policies and coal mine production capacity; ② There are significant differences in the production capacity and safety level of coal mines in different provinces of China. Through the chi square test analysis of contingency tables, the number and level of accidents in coal mine safety production are significantly correlated with the accident area. Therefore, the adjustment of coal mine production capacity in different regions is also a key factor affecting the safety situation of coal mines in China; ③ Establishing a time series prediction LSTM model that includes factors such as coal mining technology, degree of mechanization, and coal production in different regions can accurately predict the number of deaths in coal mine accidents. The correlation coefficient R² between the true and predicted values is 0.99, the root mean square error of the predicted data is 40.49; ④ According to the prediction results, while the total coal production remains unchanged, increasing production capacity in the northwest region and reducing production capacity in the northeast or southwest regions will further improve the safety production situation of coal mines; ⑤ The prediction results indicate that the number of coal mine fatalities will decrease more significantly after the production capacity shifts from southwest to northwest, but the degree of reduction in accident fatalities will decrease with the increase of time. Although the coal mining conditions in the northwest region are relatively good, it is still necessary to pay attention to the safety mining risks under the rapid development of coal production capacity. The research results can provide a reference for scientifically formulating coal industry policies.