Dia Sense-An Intelligent System for Early Diabetes Risk Prediction

Abstract

Diabetes is one of the most prevalent chronic diseases worldwide, leading to severe health complications if not detected early. Early prediction of diabetes using health indicators can significantly improve preventive healthcare. This study presents a machine learning-based approach for predicting diabetes using selected health indicators from the Behavioral Risk Factor Surveillance System (BRFSS) dataset. Key features such as BMI, age, blood pressure, cholesterol level, physical activity, dietary habits, smoking status, and general health indicators are utilized to develop predictive models. Three machine learning models—Random Forest, XGBoost, and Deep Neural Network (DNN)—are implemented and evaluated using accuracy, precision, recall, and F1-score metrics. The dataset is divided using stratified train-test splitting to maintain class distribution. Experimental results demonstrate that the Deep Neural Network model achieves superior performance compared to other models. The proposed system demonstrates the potential of machine learning techniques in assisting early diabetes detection and supporting data-driven healthcare decisions.