A New Multiclass Support Vector Machine

A new multiclass Support Vector Machine (SVM) is presented, which can be used to find the optimal decision boundaries in a multiclass classification problem. In the multiclass classification problem the goal is to construct a decision function based on a set of objects belonging to K different classes, such that the decision function best predicts the class label of a new object. The binary Support Vector Machine has been proven very successful for the classification of objects belonging to two distinct classes. The present work extends the binary Support Vector Machine to accommodate problems where objects belong to more than two classes. A single optimization problem is constructed, in which all classes are considered simultaneously. The suggested method is tested on a number of datasets, and compared with a number of other classification methods. Performance comparison is done using a bootstrap scheme which generates a confidence interval for the difference in performance between two classifiers. The different classifiers are tested on four benchmark datasets and it is shown that the proposed multiclass classification method performs at least as good as existing techniques. On one dataset the proposed method performs significantly better than all other methods. A comparison in computation time is also given, which shows that the proposed method is slower than existing methods on datasets with a large number of objects or a large number of classes. It is nonetheless believed that the proposed method provides a promising new way of looking at multiclass classification problems.