Application of binary quantum-inspired gravitational search algorithm in feature subset selection

Feature selection is an important task to improve prediction accuracy of classifiers and to decrease the problem size. Several approaches have been presented to perform feature selection using metaheuristic algorithms. In this paper, we employ the binary quantum-inspired gravitational search algorithm (BQIGSA) combined with the k-nearest neighbor classifier as a wrapper approach to select a (sub-) optimal subset of features. We evaluate the proposed approach on several well-known datasets and compare our approach with other similar state-of-the-art feature selection techniques. Comparative results verify the acceptable performance of the proposed approach in feature selection.