A SVM Model for Data Mining and Knowledge Discoverying of Mine water disasters

摘要

In order to analyze the water inrush data with a smaller number and a lower accuracy, a linear kernel H-SVMs model was presented. Firstly, a model was deduced to evaluate the generalization power of H-SVMs, then, a novel method to build H-SVMs was put forward. The separation distances of SVMs are regarded as the indices for classifying and clustering. Through the top-down and bottom-up routes, the input samples are classified by maximal separation distance and clustered by minimal separation distance. The approach of classification can select the SVM w hose separation margin is maximal through the top-down route, and dichotomize the input samples according to their categories at each node. The approach of clustering can select the SVM whose separation margin is minimal through the bottom-up route, and hierarchically cluster every two input samples according to their categories at each node. After H-SVMs' structure determined, the attributes of input samples at each SVM node is reducted, by which a closely related attributes set is constructed in order to gain a better performance for the SVM. Finally, the H-SVMs model is applied to the data mining and know ledge discoverying of mine water inrush. Experimental results show the novel method has a simple structure, and a good generalization performance, it can not only predict the scale of water inrush correctly, but also its tree structure can denote the hiberarchy of water inrush, moreover, the normal vector parameters Ws in the decision functions can describe the weights of the factors related to the mine water inrush, the prediction rules are abstracted by analyzing the decision functions, in which a novel scientific method introduced to the prediction of the water inrush.