Authors: Prabhjot Kaur; Amit Awasthi; Anchit Bijalwan
Addresses: Uttaranchal University, Dehradun, India ' Department of Physics, University of Petroleum and Energy Studies, Dehradun, Uttarakhand, India ' Arba Minch University, Arba Minch, Ethiopia
Abstract: The accuracy and performance of any machine learning model are highly dependent on the number of qualitative features taken into consideration while training the model. The selection of qualitative features depends on the considerate choice of feature selection technique. In this study, feature selection is performed using different techniques such as information gain, Gini decrease, Chi2 and FCBF on the same dataset, and subsequently, the accuracy has been measured. The results showed that the FCBF method has dramatically reduced the number of features and moderated the accuracy compared with other feature selection methods.
Keywords: feature selection; fast correlation-based feature; FCBF; network traffic; Chi2; Gini decrease; information gain.
International Journal of Computational Science and Engineering, 2021 Vol.24 No.3, pp.228 - 243
Received: 12 Apr 2020
Accepted: 10 Nov 2020
Published online: 04 Jun 2021 *