Modified reducts and their processing for nearest neighbor classification

Abstract

Dimensional reduction of data is still important as in the data processing and on the web to represent and manipulate higher dimensional data. Rough set concept developed is fundamental and useful to process higher dimensional data. Reduct in the rough set is a minimal subset of features, which has almost the same discernible power as the entire features in the higher dimensional scheme. But, we have problems of the application of reducts for the classification. Here, we develop a method which connects reducts and the nearest neighbor method to classify data with higher accuracy. To improve the classification ability of reducts, we propose a new modified reduct based on reducts and its optimization method for the classification with higher accuracy. Then, it is shown that the modified reduct improves the classification accuracy, which is followed by the optimized nearest neighbor classification. © 2012 Springer-Verlag.