A hybrid approach for feature selection

Feature selection is a process of selecting desired number of features from a large set of original features that purely contribute to the prediction of a test data with the help of a classifier. Many application areas such as, gene expression array analysis, test processing of internet document and combinatorial chemistry make use of feature selection, due to the presence of tens or hundreds of thousands of features in the dataset, the analysis of such a large number of feature set is impossible. Selecting relevant features increases the performance of predictor. Also, it makes the predictor accurate and cost effective. It also provides the better understanding of the process that classifies the data. In this paper, a new approach of feature selection has been proposed in order to improve the classification accuracy. The proposed algorithm has been tested on publicly available Heart Disease and Sonar data sets.