Selection of Affecting Factors of Coal and Gas Outburst on Genetic Algorithm

Genetic algorithm (GA) is applied to select main affecting factors of coal and gas outburst to solve the over-fitting problem of BP neural network (NN) in predicting coal and gas outburst, and a modified BP NN predictor is established, which input variables are the factors selected. In our GA, chromosome is a binary encoding which each gene corresponds to a variable, penalty function is introduced into fitness function. Finally, the method is studied using real samples of PingMei 8th mine in MATLAB2009b environment. The results demonstrate that fitting effect and prediction accuracy of the modified BP NN predictor is improved significantly and simulation time is shorter after predictor´s input valuables are optimized on GA.