Leveraging Diabetes Prediction using the Deep Learning-based Hybrid ANN-CNN Architecture

Abstract

Diabetes mellitus is a worldwide pandemic chronic metabolic disease that threatens human health seriously. Correct and early prediction of diabetes is one of the important factors for medical treatment and diabetes management. In the meanwhile, this study, proposed a deep learning-based framework as a novel method of prediction of diabetes using a large-scale dataset containing over 6000 patient records. This study reviews several deep learning algorithms, such as, Artificial Neural Networks (ANN) and Convolutional Neural Networks (CNN) to investigate the most effective algorithm for diabetes prediction. This study formed hybrid architecture ANN-CNN by utilizing the strengths of each model. We perform extensive preprocessing of the data and extract features that contribute to the models' efficiency. The experimental results of this study demonstrated that hybrid ANN-CNN architecture model achieved the highest accuracy of 94.3% clearly outperforming standalone ANN (89.2%) and CNN (91.4%) models in prediction accuracy, showcasing their potential in clinical decision support systems.