COMPARATIVE ANALYSIS OF EVOLUTIONARY TECHNIQUES FOR FEATURE SELECTION AND CLASSIFICATION

Abstract

Classification problem is the main issues in the large number of features data sets but not all classification algorithms are useful. Here data reduction and feature selection done with the help of (normalize by subtracting mean, covariance, eigenvectors, eigenvalues, significant Principal Components). It reduced the redundant and irrelevant features from the data sets. Feature selection and classification is done by Artificial Neural Networks, its select the relevant features to achieve the better classification performance and reducing the number of feature. This report presents the one comparative study based on Particle Swam Optimization (PSO), Ant Colony Optimization (ACO) and Multiple Objective Particle Swam Optimization (MOPSO) on the train data which in generated by the classification. We investigated PSO-based multi-objective algorithm performed based as compared to single objective PSO and ACO. The MOPSO algorithm presents idea based on non-dominated arrangement into PSO for address the feature selection problems. It accomplished the equivalent results with other two well-known algorithms in most cases.