Minimum Redundancy Maximum Relevance(mRMR) Based Feature Selection Technique for Pattern Classification System

Abstract

Feature Selection is an important hurdle in classification systems. We study how to select good features by making the covariance matrix of each sample data set and extracting the features from it .Then, we try to find out the length of each sample by finding the error rate .We perform experimental comparison of our algorithm and other methods using two data sets(binary and functional) and three different classifiers(support vector machine, linear discriminant analysis and naïve Bayes).The results show that the MRMR features are less correlated with each other as compared to other methods and hence improves the classification accuracy.