Leveraging Hybrid Ensemble Framework for Diabetes Prediction: Integrating Random Forest, ANN, ETC, and Stack-CNN

Mellitus (DM), or diabetes, is a dominant universal medical affair dilemma harming a growing number of people around the world and putting them at severe risk. Diabetes is a long-lasting affliction that leads to a high annual mortality rate. Diagnosing diabetes initially is crucial to limit its growth and reduce the risk of serious kidney disease and eye disease. This paper recommends using an ensemble hybrid framework for a high-accuracy DM prediction. In this paper, a hybrid ensemble framework blends ML and DL the machine and deep learning structures with an ensemble learning-based stacking model to improve diabetes prediction. The Diabetes Dataset from the Frankfurt Hospital Germany (DDFH-G) trains the hybrid model. The conduct of the recommended model was assessed based on numerous parameters like accuracy, precision, specificity, sensitivity, ROC/AUC, F-score, and Matthews Correlation Coefficient. The recommended hybrid framework with stack-CNN achieves excellent performance with a 99.24% accuracy rate in testing the Diabetes Dataset from Frankfurt Hospital (DDFH-G). The results show that the recommended hybrid framework outperformed previous studies on independent machine and deep learning structures in the earliest diabetes screening.