Ensemble Method Using Correlation-Based Feature Selection with Stratified Sampling for Classification

Abstract

Ensemble methods are preferred as they represent good significance over specific predictor regarding accuracy and confidence in classification. This paper proposes here the ensemble method with multiple independent feature subsets in order to classify high-dimensional data in the area of the biomedicine using Correlation feature selection with Stratified Sampling and Radial Basis Functions Neural Network. First, the method selects the feature subsets using Correlation-based feature Selection with Stratified Sampling. It minimizes the redundancy in the features. After generating the feature subsets, each feature subset is trained using base classifier and then these results are combined using majority voting. The proposed method uses CFS-SS in ensemble classification method.