Sami Kaya

Overlap Mitigation Versus Classifier Selection in Imbalanced Classification: A Dual-Baseline Analysis

A full factorial experiment was run across 58 KEEL datasets, 11 mitigation techniques, and 10 classifiers, suggesting that under typical practitioner conditions, a strong-classifier default is the lower-effort, lower-variance path, with no expected performance cost relative to a pipeline-development investment.