Fig. 10
From: An approach for handling imbalanced datasets using borderline shifting
Mean Results obtained by the SVM classifier using different resampling techniques across all benchmark datasets. Each bar (or point) represents the average classification performance achieved after applying a specific oversampling or hybrid method. The results indicate that the proposed Borderline-Shifting Oversampling (BSO) method consistently yields higher AUC values compared to existing techniques, confirming its effectiveness in improving minority-class recognition and reducing class overlap.
