Approach to verify new class in the classification process

Abstract

Support vector machine (SVM) is a binary class machine learning classifier. Given a data point, the SVM can classify the data point to either positive class or negative class. However, in some cases, some data points belong to neither positive class nor negative class. They should be treated as one new class. This paper proposes one method that can find isolated data points and separate them into new classes based on F-test and the experimental results show that the method is effective.