Abstract: A method based on the rough set neural network(RSNN)for the prediction of the coal mine spontaneous combustion is presented in this paper.The measured data is decreased in this way by use of the rough set reduction theory,the data is based on the intensity of the wind leak Q and the temperature of the coal mine T_C measured in the mined-out area of the fully mechanized long-wall top-coal caving face.Then the RSNN is established on foundation of the data reduced,and the minimum thickness of the mine layer is predicted using the RSNN.The real-time measured data shows that this method is simpler than the ordinary AMAX prediction method and its precision is high.The method lays a good foundation for the network-based remote coal mine safety monitoring and control system.