An Efficient Hybrid Algorithm For Multi - Class Support Vector Machines

The standard support vector machines (SVM) algorithm is originally designed for two-class classification. it has been applied to solve multi-class classification problems. Several algorithms are developed for solving a multi-class problem by SVM such as one-against-one (OAO), one-against-all (OAA), and directed acyclic graph support vector machines (DAGSVM). In this research, a hybrid algorithm for solving a multi-class classification problem by SVM is proposed. In this algorithm, all classes are divided into two major classes based on the distance among center point of classes and range of spread data in those classes. In order to compare the efficiency and the accuracy of the proposed algorithm, the proposed algorithm and the OAO, OAA, and DAGSVM algorithms are applied to four datasets. The results show that the efficiency and the accuracy of the proposed algorithm are better than the others. The proposed algorithm also solves unclassifiable region problem that may exist in OAO and OAA algorithms. Keywords—Center Point, Multi-Class Classification, Spread Data, Support Vector Machines