Based on the comprehensive study of the various algorithms, with the rough set theory and the Least Squares Support Vector Regression, taking advantage of rough set method can remove redundant information, Least Squares Support Vector Regression can accurately accelerate the convergence speed advantages. Prominent use of a specific network prediction mechanism, and use this prediction of mine gas outburst mechanism to predict the situation. After based on the MATLAB neural network toolbox BP neural network method of experimental comparison shows that the LS-SVR can speed up the convergence rate. The experimental result reveals that Based on RS-LS-SVR neural network prediction model is reliable, fast convergence and high accuracy, good effect.