No abstract

No abstract

No abstract

Oversampling methods are among the most widely used methods to solve imbalanced data analysis due to their simplicity and flexibility. Multi-class imbalanced problems involve critical issues related to synthetic samples and corresponding consequences on the classification results. The over-fitting phenomenon, the overlapping regions between classes, and the existence of noisy original samples may lead to a non-optimal data distribution, which inadequately affects the classifier performance.To avoid these limitations, we propose a novel oversampling technique based on the intra-distance-matrix (IntraDM) and the inter-distance-matrix (InterDM). The new samples are created to decrease over-fitting, decline the generation of outliers, and minimize the overlap areas. Applying the proposed method on 14 available datasets shows that it outperforms common oversampling techniques in terms of classification evaluation metrics.

No abstract