Emphatic Constraints Support Vector Machines for Multi-class Classification
Mostafa Sabzekar, Mahmoud Naghibzadeh, Hadi Sadoghi Yazdi, Sohrab Effati · 2009
Support vector machine (SVM) formulation has been originally developed for binary classification problems. Finding the direct formulation for multi-class case is not easy but still an on-going research issue. This paper presents a novel approach for multi-class SVM by modifying the training phase of the SVM. First, we propose the Emphatic Constraints Support Vector Machines (ECSVM) as a new powerful classification method. Then, we extend our method to find efficient multi-class classifiers. We evaluate the performance of the proposed scheme by means of real world data sets. The obtained results show the superiority of our method.