PHGA: Proposed hybrid genetic algorithm for feature selection in binary classification

Abstract

In data pre-processing, feature selection has particular importance. Selection of appropriate features in classification leads to accuracy enhancement, reduction of execution time and increment of model interpretability. In this paper, an innovative algorithm for feature selection is proposed which combines filter and wrapper techniques and uses benefits of evolutionary computation and assembly of independent measures. Experimental results show the superiority of the proposed hybrid genetic algorithm for feature selection in binary classification (PHGA). In comparison with other existing methods PHGA can select the smaller set of features in a shorter execution time with higher classification accuracy.