Development of a Neighborhood Based Adaptive Heterogeneous Oversampling Ensemble Classifier for Imbalanced Binary Class Datasets

Class imbalance prevails in many real-word datasets. In this paper, a Neighborhood based Adaptive Heterogeneous Oversampling Ensemble Classifier method is proposed to handle class imbalance in datasets. The proposed method adopts an oversampling approach to create a set of balanced representative training datasets. Several base classifiers are built based on those training datasets, and an adaptive heterogeneous ensemble classifier is created. The proposed method is examined with five datasets, and examination results are compared with popular oversampling algorithms. The comparison revealed that proposed method is able to achieve better performance results.