DE/EI - A New Differential Evolution Selection Operator Based on Entropy and Index for Feature Ranking

Abstract

The goal of feature ranking is to find the optimal list of features. The feature ranking methods use different search techniques to select features. An optimal feature selection results in an optimal feature ranking list, so, it is necessary to use a stochastic search method to select features. In this article, a new DE selection operator is introduced. To know the value of the features its fitness function is calculated using Shannon and singular value decomposition entropy. The index of the selected feature is computed by JI, MCI, and AA Index to know the feature list stability. Hence, DE/EI parent selection operator is proposed. The six fitness functions: SMCI, SVDMCI, SJI,SVDJI, SAA, and SVDAA, are thoroughly tested on ten UCI data sets and their performance is measured with different classifiers like Naive Bayes and Support Vector Machine. The experimental results show that the proposed method can efficiently be consolidated into any evolution that is based on a parent selection framework.