Enhancing Prediabetes Prediction Using Hybrid CNNs and SMOTE for Data Imbalance

Prediabetes, a condition characterized by elevated blood glucose levels, is a critical precursor to Type II diabetes. Early detection of prediabetes can significantly reduce the risk of progression to full-blown diabetes, as well as associated complications. However, the imbalance in medical datasets poses a challenge to accurate prediction, often leading to the underper-formance of conventional machine learning models. In this study, we propose a hybrid Convolutional Neural Network (CNN) model combined with the Synthetic Minority Over-sampling Technique (SMOTE) to enhance the prediction of prediabetes. By addressing the class imbalance problem, our approach improves the predictive accuracy for the prediabetic class, which is typically underrepresented in datasets. The hybrid CNN model is evaluated against traditional machine learning techniques, such as logistic regression, decision trees, support vector machine (SVM) and artificial neural network (ANN), demonstrating superior performance in terms of precision, recall, and overall accuracy. Our results indicate that integrating SMOTE with deep learning models offers a robust solution for enhancing the early detection of prediabetes, providing healthcare practitioners with a valuable tool for implementing timely preventive interventions.