Unsupervised hybrid PSO — Relative reduct approach for feature reduction

摘要: Feature reduction selects more informative features and reduces the dimensionality of a database by removing irrelevant features. Selecting in unsupervised learning scenarios is harder problem than supervised feature selection due to absence class labels that would guide search for relevant Rough set proved be efficient tool needs no additional information. PSO (Particle Swarm Optimization) an evolutionary computation technique which finds global optimum solution many applications. This work combines benefits both rough sets better data reduction. paper describes novel Unsupervised based Relative Reduct (US-PSO-RR) employs population particles existing within multi-dimensional space dependency measure. The performance proposed algorithm compared with methods USQR (UnSupervised Quick Reduct) USSR effectiveness approach measured using Clustering evaluation indices.