A new method to feature selection in highdimantional data sets for data mining

Abstract

A method of feature selection by using clustering graph is presented. in the suggested method, using a detection algorithm communities at first, the basic characteristics divided into some clusters. Then, by applying genetic algorithms and using a KNN (K nearest neighbor) classification algorithm a feature selection method based on covering solution is presented. The performance of the suggested method compared with the most recognized and the most recent feature selection methods applied on different classificators. The results showed that the suggested method both in terms of time and classification accuracy has a proper function.