A New Multi-class Classification Method Based On Least Square Support Vector Regression Machine

The support vector machine is a commonly used classification method for its good performance. The processes of most of the existing methods for multi-class problem are not simple. More than one support vector machine (SVM) classifier should be trained in each of these methods. In this paper, a novel multi-class support vector machine method is presented, named the Multi-class Least Square Support Vector Regression (MCLS-SVR) by using only one classifier to solve the multi-class problem. The main idea of this method is to transform the classification problem to the regression problem, which is much simpler, by treating the class label of each sample as the regression output value; then LS-SVR is used to solve the regression problem and a rounding operation is added in the course of testing samples because their class labels are positive integers. The results of both simulation and experimental on face and fingerprint recognition show that the proposed MCLS-SVR method yields higher recognition rate and faster calculation speed.