Fig. 5
From: An approach for handling imbalanced datasets using borderline shifting
Workflow of the proposed BSO algorithm. Phase 1 identifies borderline majority-class instances near the decision boundary (red outline). Phase 2a reclassifies selected borderline instances as minority class (green arrows indicate class change). Phase 2b performs targeted resampling to generate synthetic minority instances (pink arrows) and balance the dataset. The diagram illustrates both the identification and enhancement of borderline instances to improve classifier performance.
