Class imbalance problem using a hybrid ensemble approach

Article type: Research Article

Authors: Abd Elrahman, Shaza M.^a | Abraham, Ajith^{b; *}

Affiliations: [a] Faculty of Computer Science & Information Technology, Sudan University of Science Technology, Khartoum, Sudan | [b] Machine Intelligence Research Labs, Scientific Network for Innovation and Research Excellence, Auburn, WA, USA

Correspondence: [*] Corresponding author: Ajith Abraham, Machine Intelligence Research Labs (MIR Labs), Scientific Network for Innovation and Research Excellence, Auburn, WA, USA. E-mail:ajith.abraham@ieee.org

Abstract: This paper proposes a comparative study that investigates the effects of using resampling (undersampling and oversampling) methods with homogenous ensemble methods Bagging and AdaBoost in imbalanced data sets. We presented a hybrid ensemble approach that combined multi resampling by integrating both undersampling and oversampling to get benefits and reduces drawbacks caused by each of them. The proposed approach has improved the performance even those most sensitive to imbalanced class data sets.

Keywords: Data mining, machine learning, class imbalance, ensemble learning

DOI: 10.3233/HIS-160217

Journal: International Journal of Hybrid Intelligent Systems, vol. 12, no. 4, pp. 219-227, 2015