Answers are grounded in this paper and cite their sources.
Diabetes Mellitus (DM) is a major chronic disease whose early prediction is hindered by complex nonlinear clinical relationships and class imbalance in medical datasets. Existing ensemble and non-ensemble Machine Learning (ML) methods are often degraded by this imbalance, especially for minority-class patients. To Address this problem, this study proposes an adaptive ensemble learning framework that combines the Synthetic Minority Oversampling Technique (SMOTE), Reinforcement Learning (RL), the Bat algorithm, and an AdaBoost-based fitness function to construct an optimized ensemble. SMOTE first balances the training data by generating synthetic minority samples. Then an RL-Driven Bat algorithm explores the ensemble search space to find an optimal set of base classifiers, while the AdaBoost-based fitness function evaluates candidate ensembles and drives performance maximization. The proposed model evaluates on the PIMA Indians Diabetes (PID) dataset, the framework attains 0.95 accuracy and 0.98 specificity. Further, to validate the significance of the proposed framework, a Wilcoxon Signed-Rank (WSR) test is conducted between the proposed model and other ensemble and non-ensemble classifiers. This test shows that p < 0.05 for all traditional ensemble and non-ensemble models, indicating the statistical significance of the proposed model. An ablation study quantifies the individual impact of SMOTE, RL, the Bat algorithm, and the AdaBoost-based fitness mechanism. Comparative results against state-of-the-art (SOTA) ensemble, non-ensemble, and metaheuristic-based ensembles show superior accuracy and specificity, indicating that the proposed adaptive optimization-based ensemble is a robust and effective solution for early DM prediction under class-imbalanced conditions.
No takes yet. Share an insight, caveat, or question.
Eedunuri et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: