Intelligent Hybrid Approach for Feature Selection

Abstract

The issues of multitude of noisy, irrelevant, misleading features, and the capability to tackle inaccurate and inconsistent data in real world topics are the justification to turn into one of the most significant needs for feature selection. This paper proposes an intelligent hybrid approach using Rough Set Theory (RST), Chaos Theory and Binary Grey Wolf Optimization Algorithm (CBGWO) for feature selection problems. Ten different chaotic maps are used to estimate and tune GWO parameters. Experiments are applied on complex medical datasets with various uncertainty features and missing values. The performance of the proposed approach is extensively examined and compared with that of existent feature selection algorithms; such as ant lion optimization (ALO), chaotic ant lion optimization (CALO), bat optimization (BAT), whale optimization algorithm (WOA), chaotic whale optimization algorithm (CWOA), binary crow search algorithm (BCSA), and chaotic binary crow search algorithm (BCCSA) algorithms. The achievement of the proposed approach is analyzed using different evaluation criteria. The overall result indicates that the proposed approach delivers better performance, lower error, higher speed and shorter execution time.