Abstract: In order to improve the accuracy of Fuzzy C-Means Clustering Algorithm (FCM) in prediction of coal and gas outburst, a method was putted forward. The method uses fuzzy C-means clustering that combines simulated annealing algorithm (SA) and genetic algorithm (GA) to predict coal and gas outburst. The method combines global search, high-precision advantages of simulated annealing algorithm and powerful search ability of space of genetic algorithm. Initial value optimized by the genetic simulated annealing algorithm is assigned to FCM. It avoids FCM algorithm convergence to local minima when the initial cluster center value is not properly selected in combination with the analysis of typical outburst mine data, the results indicate that FCM optimized by the genetic simulated annealing algorithm is more accurate than the single FCM algorithm in prediction of coal and gas outburst.