Abstract:Aiming at the problem that the classification performance of binary trees is greatly affected by its hierarchical structure,this paper proposes an improved multi-classification algorithm for partial binary tree twin support vector machine(TWSVM).The algorithm defines a mixed separation measure β,which based on sample class distance and interclass distance,and a suitable binary tree twin support vector machine classifier is constructed based on the value of β.Compared with other SVM multi-classification algorithm on UCI data sets,the superiority of improved algorithm is verified.Taking the wind turbine gearbox as the research object,feature vectors of faults were extracted based on time-frequency joint feature,and the faults were diagnosed by the improved partial binary tree TWSVM.