Research on the Prediction Method of Monthly Hidden Danger Quantity in Coal Mine Based on BP Neural Network Periodic Combination Model

To better prevent the occurrence of hidden dangers coal mine accidents and ensure safety production enterprises. This paper mines analyses pattern historical monthly danger quantity in constructs three models: traditional backpropagation (BP) neural network model, BP based on adaptive moment estimation optimization algorithm (Adam-BP) prediction model with introduction moderators (Month-Adam-BP). The experimental results show that Adam-BP can improve accuracy, which mean absolute percentage error (MAPE) improves by 8.93%, root square (RMSE) 8.15%, postdifference ratio C 0.04, small probability P 0.12; Month-Adam-BP adjustment factor further MAPE 2.61%, RMSE 5.41%, 0.06, 0.03. And accuracy reaches level 2 standard credible effect; it also be used to predict periodic characteristics hazard data. Our predicted number hazards this for next month is 29, an increase compared previous month. Thus, managers need strengthen management accidents, serve standardization smooth mine.