Feature Selection by Using DE Algorithm and k-NN Classifier

Abstract

Feature selection (FS) in data mining is one of the most important preprocessing steps before starting analysis processes. In addition to more accurate estimation of the results by the FS process, unnecessary resource utilization can be avoided by removing unnecessary attributes. One of the most commonly used data prediction and classification methods in data mining is k Nearest Neighbor algorithm (k-NN). k-NN classifier can classify data swiftly. In this study, FS process with Differential Evolution (DE) algorithm has been realized by using sample datasets. During the cost accounting of candidate solutions of DE algorithm, k-NN classifier has been used. The results obtained with DE algorithm have been compared to Correlation based feature selection, Information gain based feature selection and Learner based feature selection methods that are commonly used in classification problems. The results showed that the proposed method could successfully perform FS process.