Minimum Spanning Tree Support Vector Machine Based on Fisher Separability Measure

Classification strategy is an important issue of support vector machine application. Aiming at the defects of common used classification methods, an improved minimum spanning tree support vector machine (MST-SVM) is proposed. MST-SVM has the advantages of simple structure and high classification efficiency. The classification process is further optimized by introduction of Fisher separability measure in feature space, and the classification performance is improved. The construction process of MST-SVM is given in this paper, the effectiveness and generality of the new method are proved by contrast experiment on the standard data sets, and the application value is proved by analog circuit fault diagnosis.