Diabetes Risk Prediction Using Classical Machine Learning Models and Flask-Based Web Deployment

Diabetes mellitus is a chronic metabolic disorder and a major public health concern worldwide. Early prediction of diabetes risk from routinely collected clinical attributes can support timely intervention and reduce long-term complications. This paper presents DiabetesML, a comparative study of classical machine learning models for diabetes risk prediction using the Pima Indians Diabetes Dataset. The system evaluates Logistic Regression, Decision Tree, and Random Forest classifiers on eight clinical features including pregnancies, glucose level, blood pressure, skin thickness, insulin, body mass index (BMI), diabetes pedigree function, and age. Logistic Regression achieves the highest test accuracy of 78.57%, while Random Forest and Decision Tree achieve 75.97% and 69.48%, respectively, on a stratified 80–20 train–test split. In addition to accuracy, models are assessed using precision, recall, F1-score, ROC-AUC, and confusion matrices to provide a comprehensive performance anal-ysis. The best-performing model is deployed via a Flask-based web application with an HTML frontend, enabling end-users to submit clinical parameters and receive real-time diabetes risk predictions. The proposed system demonstrates the practicality of integrating classical machine learning with lightweight web technologies for accessible clinical decision support.