Optimal feature selection using elitist genetic algorithm

Abstract

A method of feature selection using elitist Genetic Algorithm is proposed in this work. Stratified-tenfold-cross-validation classification accuracy is used as fitness function. The method developed can detect redundant and irrelevant features, consequently producing the optimal feature set. The algorithm is carried out on the four benchmark datasets. Results of the experiments carried out shows that the algorithm developed selects the best set of features in terms of stratified-tenfold-cross-validation classification accuracy. Finally, the results obtained are compared with established results for the same datasets. Improvement in the size of selected subsets are also demonstrated. © 2015 IEEE.