A Novel Stacked Ensemble Approach for Diabetes Prediction: Merging Machine Learning and Deep Learning Techniques

To develop an intelligent AI-based predictive hybrid model that accurately identifies diabetic by analyzing clinical and demographic data for early detection and improved healthcare decisionmaking. This study introduces a hybrid model combining machine learning (ML) and deep learning (DL) techniques for enhanced diabetes prediction. By stacking three ML models (Random Forest, XGBoost, and Logistic Regression) and three DL models (CNN, FNN, and DNN), followed by soft voting, the research aims to leverage the strengths of both approaches to improve accuracy, recall, and precision. The model was tested on the Pima Indians Diabetes Dataset and the LMCH dataset, demonstrating superior performance across key metrics, particularly in handling imbalanced datasets. The hybrid model achieved the best results with an accuracy 92.48%, precision 97.64%, Recall 87.07%, F1 Score 92.05%, ROC-AUC 92.48%, Cohen's Kappa 84.96%, making it a promising tool for early diabetes detection. A Hybrid AI model is superior to XGBoost, CNN for diabetes prediction because it synergistically combines multiple learning paradigms to achieve deeper feature representation, higher predictive accuracy, and stronger clinical reliability. Additionally, the study emphasizes the importance of integrating ML and DL models to improve generalization and robustness in medical predictions.