An Integrated Optimization Technique with SVM for Feature Selection

This manuscript proposes an integrated optimization technique of binary particle swarm optimization (BPSO) with binary successive approximation (BSA) technique. The proposed technique along with support vector machine (SVM) is implemented to solve the feature selection problem by excluding redundant features to improve the classification accuracy. Feature selection constitutes among the most challenging problems in data mining and machine learning. In the suggested method, the BSA method is used to further improve the result produced by the BPSO method. The evolutionary search (ES) strategy is the foundation of the BSA process in which exploration is carried around a hypercube, which is computationally efficient and able to search for better solutions in proximity. The proposed technique is tested on four datasets from UCI, Kent Ridge, and the Gene Expression Model database. The obtained results were proved to be acceptable when compared to those that had been published.