Mine gas content prediction model is a multivariable nonlinear function relation, the accurate prediction model is established in various influence factors depends on the interaction between the mutual coupling. The neural network and the particle swarm algorithm organically, based on neural network theory, using particle swarm optimization algorithm and the number of hidden neurons in the network connection weights, gas content prediction model is established. Solved the bp neural network, slow convergence speed, easy in local optimum. And according to the historical data, establishing genetic neural network training and testing samples, and use of matlab simulation, the results show that particle swarm neural network model reliability, high precision.