A novel ensemble feature selection method through Type I fuzzy

Abstract

These days, one of the needed methods in machine learning is feature selection. In other words, in this manner, the most fitting features are picked. Nevertheless, there are various feature selection methods, getting the most suitable features still is a complex problem. Lately, applying several feature selection methods rather than a unique feature selection method is more efficient. In this paper, a new ensemble feature selection method based upon fuzzy Type-I named EFSF is presented. First, three different individual feature selection methods are applied to determine the rank of features separately. Next, Type-I fuzzy handles feature selections' uncertainty and decrease noise to give each feature the best rank. To validate the act of EFSF, it is compared with some ensemble methods and several advanced feature selection methods. EFSF is assessed based on Accuracy, Precision and Recall, metrics. The outcomes verify that the EFSF is better than its competitors. The source code of EFSF is here.