DIMENSIONALITY REDUCTION: ROUGH SET BASED FEATURE REDUCTION

Abstract

Feature selection refers to the problem of selecting those input features that are most predictive of a given outcome; a problem encountered in many areas such as machine learning, pattern recognition and image processing. In particular, this has found successful application in tasks that involve datasets containing huge numbers of features which would be impossible to process further. Recent examples include cluster analysis and image classification. Rough set theory has been used as such a dataset pre processor with much success, but current methods are inadequate at tending minimal reductions. This paper proposes a new feature
selection mechanism based on fuzzy forward and backward reduct. It also presents a new entropy- based modification of the original rough set-based approach. These are applied to the problem finding minimal rough set reducts, and evaluated experimentally.