Abstract: Gas content in coal seam is one of important performance indexes of safety production in coal mine,but routine predicting methods based on experience and traditional mathematical model are difficult to predict gas content in coal seam accurately.Aiming at the problem,on base of analyzing model of improved BP neural network based on Fletcher-Reeves conjugate gradient method,combining with kinds of influence factors of gas content in coal seam,the paper established a predicting model of gas content based on improved three-layer BP neural network and did concrete network training and predicting simulation.The results showed that the predicting model of gas content has quick convergence speed and high predicting precision,which can meet requirements of practical production.