An effective supervised filter based feature selection algorithm using rough set theory

Data is generally represented by high dimensional feature vectors in many areas, such as pattern recognition, data mining and machine learning. Classification of useful knowledge in high dimensional data collections is an important and demanding area. Rough set theory, is a significant component of soft computing paradigm for data analysis based on classification of objects of interest into similarity classes, which are indiscernible with respect to some features. This theory offers fundamental concepts of attribute (feature) reduction. In this work supervised feature selection algorithms using Rough set theory which falls under filter method is studied. An enhanced version of Rough set theory based algorithm is proposed which exploits the lower approximation, dependency and significance measure of attributes. The experimental analysis for the proposed method is performed on five data sets of UCI machine learning repository. The performance of the reduced data set is measured by the classification accuracy and it is evaluated using WEKA classifier tool. Result analysis and comparison shows the efficiency of the proposed algorithm.