Application study of data mining method in the prediction of mine water inrush

A new nonlinear method is proposed for predicting the mine water inrush. According to the fractal theory, the phase space of time series obtained from the mine water inrush is reconstructed. And the minimum embedding dimension is determined with Cao method, then the minimum embedding dimension is used for the input node of the support vector machines. The prediction model of time series is established based on the support vector machines. The mine water inrush of new well in Fang Shan mine is predicted with this method. And the prediction results show that they coincide with the observing values better. Different dimensions is chosen to predict the mine water inrush, and when the minimum embedding dimension is four, its prediction precision is the highest. Comparing with the predicting results of three methods including least square method, index fitting method and SVM method, the prediction precision of SVM is the best. It proves that the method put forward in this paper is reliable to predict the mine water inrush.