On Multiclass Classification Methods for Support Vector Machines

The multiclass SVM methods based on binary tree are proposed. The new methods can resolve the unclassifiable region problems in the conventional multiclass SVM methods. To maintain high generalization ability, the most widespread class should be separated at the upper nodes of a binary tree. Hypercuboid and hypersphere class least covers are used to be rules of constructing binary tree. Numerical experiment results show that the multiclass SVM methods are suitable for practical use.