Improving Computational Performance of Least Squares Multiple Birth Support Vector Machines with k-Means Clustering

Multiple Birth Support Vector Machines (MB-SVM) were introduced as a powerful extension of SVM. Although the basic idea is like SVM, optimal non-parallel hyperplanes are used for each class category. One may find different implementations of this approach in the literature. One of these is the Least Squares MB-SVM. An appealing property of this implementation is that an analytical solution is obtained that is used for classification. On the other hand, this solution involves matrix computations of sizes depending on the number of attributes and the size of the data set. In this study, we propose to use the k-means clustering algorithm before applying the Least Squares MB-SVM algorithm to improve the computational performance of MB-SVM. The preliminary results with the Iris data set indicate that by using only a small number of examples obtained from the k-means clusters, comparable performance can be obtained with the LS-MB-SVM method.