Sub-Feature Selection for Novel Classification

Abstract

Feature selection has emphasized in data mining research field on several data sets such as hospital data, text data, finance data etc. Most of the feature selection methods have been applied on traditional features for classification. The existing class can't solve the real world problem every day. Thus, it needs to generate the new class from existing class to solve several classification problems. With the help of tiny features value (especially sub-feature values), the new class has generated to find appropriate solution with reason from existing database using several statistical and optimization method like simple probability, Lagrangian function and feature selection method. This paper proposed the sub-feature selection framework to identify the distinguish class from traditional class with effectiveness. The experimental results of this work reveal the distinctness of novel class and identified the sub-feature data towards new class.