When Does Hyperparameter Optimization Matter in Imbalanced Medical Classification? A Systematic Benchmark Study Across Cost Asymmetry and Imbalance Conditions

Class imbalance and asymmetric misclassification costs are pervasive in medical machine learning, yet their joint interaction with hyperparameter optimization (HPO) method selection remains largely uncharacterized. We present the first systematic benchmark study quantifying under which combinations of imbalance ratio (IR) and clinical cost asymmetry (κ) the choice of HPO method has meaningful impact on diagnostic performance