Diabetes Prediction Using Convolutional Neural Network and Random Forest

Diabetes, a prevalent global ailment, poses a significant threat to human health, emphasizing the urgent need for early detection to impede its progression.This study presents a novel approach utilizing deep learning for early diabetes detection.Leveraging the PIMA dataset, which comprises numerical attributes common in medical data, we address the limitations of conventional convolutional neural network (CNN) models by transforming numerical features into images based on their importance.Subsequently, three classification strategies are employed: direct classification using ResNet18 and ResNet50 CNN models, fusion of deep features from ResNet models followed by support vector machine (SVM) classification, and direct classification of selected fusion features by SVM.Results underscore the efficacy of diabetes images in facilitating early diagnosis, signifying a promising avenue for enhancing healthcare outcomes.