Abstract: We acquired the correlation coefficients between coal seam gas permeability and its influential factors which are effective stress, temperature, compressive strength and gas pressure by correlation analysis. In addition, the influence factors of coal seam gas permeability have cross correlation among each other according to correlation analysis. The mean impact value （ＭＩＶ） method was used for the optimization of influencing factors and achieved the three main influential factors affecting coal seam gas permeability which are effective stress, temperature and gas pressure, and choosing them as input variables of BP neural network for modeling based on experimental data. In this study, two models of coal seam gas permeability were established, one of them is called model 1 which was built without optimization of influencing factors and the other one is called model 2 which was built with optimization of influencing factors. Through the modeling computation and error analysis, we can safely conclude that the model 2 has better stability and higher accuracy in model predictive, and it can reflect the mapping relationship between coal seam gas permeability and its influencing factors perfectly.