A comparative study on the effect of feature selection on classification accuracy

Feature selection has become interest to many research areas which deal with machine learning and data mining, because it provides the classifiers to be fast, cost-effective, and more accurate. In this paper the effect of feature selection on the accuracy of NaiveBayes, Artificial Neural Network as Multilayer Perceptron, and J48 decision tree classifiers is presented. These classifiers are compared with fifteen real datasets which are pre-processed with feature selection methods. Up to 15.55% improvement in classification accuracy is observed, and Multilayer Perceptron appears to be the most sensitive classifier to feature selection.