A Hybrid γ-ABC-Feature Selection Approach for Improving Disease Classification

This paper proposes a γ-Artificial Bee Colony – Feature Selection (γ-ABC-FS) approach to identify the salient feature subset that improves classification accuracy. γ-ABC-FS hybridizes Artificial Colony(ABC) algorithm by employing Rough Set Theory concepts in ABC search process and improved initialization local strategies. The proposed is intended begin evolutionary not excluding attributes contribute specific decision initial solution set. And designed hybrid strategy considers number of selected features, their dependency degree, accuracy as significant parameters yield an optimal set with better convergence performances. Experiments on UCI disease datasets revealed remarkably generated subsets appreciable improvement outperformed state art approaches all aspects.