C. Navaneethan

Quantifying the imputation paradox and XAI inconsistency in multi-source diabetes prediction: a 353,680-record leakage-free stacking ensemble with dynamic routing architecture

Diabetes affects 537 million adults globally, a figure projected to reach 783 million by 2045. Despite over 4,200 ML prediction studies, clinical translation remains hindered by an over-reliance on benchmark datasets, unmeasured information costs of multi-source fusion, and untested XAI convergence assumptions. We addr...