A Novel Method of Sparse Least Squares Support Vector Machines in Class Empirical Feature Space

In this paper, we propose a novel method of sparse least squares support vector machine (SLS-SVM) that is trained in each class empirical feature space spanned by the independent training data belonging to the associated class. And we determine the decision function in each class empirical feature space. To prevent that the information of other classes is lost because of generating each class empirical feature space separately, we combine the decision functions of all the classes by training LS-SVM in primal form. Using benchmark data sets, we evaluate the effectiveness of the proposed method over the conventional methods.