Binary Tree SVM Based on Analytic Hierarchy Process and Its Application to Fault Diagnosis

Abstract

Structure design has a great influence on classification accuracy of the binary tree support vector machines, to design the structure reasonably, a multi-class algorithm of the binary tree SVM based on analytic hierarchy process is proposed. Based on analytic hierarchy process (AHP), the model of an assessment system is established. First combine theoretical analysis with expert advice, make a comprehensive survey on several factors and confirm the weight of faults, then put the faults in the right order based on the weight, last the structure of the binary tree is designed. It is proved by the fault diagnosis of a motor that the method runs well in classification accuracy and generalization ability and is suitable for multi-class.