Abstract: In order to forecast coal gas permeability effectively，warning gas density variation underground，self adaption neural fuzzy inference system(ANFIS) forecast model was conducted by self adaption studying ability of neural network and experiential knowledge of fuzzy inference system，based on experimental data，the forecast results and the results from BP neural network model and SVM model was compared.The results showed that convergence rate was quickly for ANFIS model，forecast value approached practical value better，it’s performances were better than other two models from error precision，training speed and convergence rate and so on.The high precision forecast of gas permeability could be forecast form effective stress，gas pressure，temperature and compressive strength.