Integrating Rough Set Theory and Particle Swarm Optimisation in feature selection

This paper proposes a new feature-selection strategy by integrating the Rough Set Theory (RST) and Particle Swarm Optimisation (PSO) algorithms to generate a set of discriminatory features for the classification problem. The proposed method is seen as a marriage between filter and wrapper approaches in which the RST is used to pre-reduce the feature set before optimisation by PSO, a meta-heuristic approach using Support Vector Machines (SVMs). Experimental results, based on the number of reducts and classification accuracy, were compared for the grid search method using data from the Machine Learning Repository. For most datasets, the proposed method statistically significantly improves the obtained classification accuracy and reduces the number of feature subsets.