Diabetes Risk Predictor: Machine Learning-Based Early Diagnosis is a comprehensive project that implements a robust machine learning pipeline for early diabetes risk prediction. This project is based on the methodology and findings of the research paper titled “Diabetes Diagnosis Using Machine Learning: High-Accuracy Predictive Modeling with Diverse Data Sources and Methods” by Jasmin Patel and colleagues. The project features an interactive Streamlit web app that provides a user-friendly interface for risk prediction and experimentation.

The app supports multiple datasets, including the Pima Indians Diabetes Database and a Frankfurt hospital dataset, and allows users to upload their own data. Users can choose from various imputation and scaling methods for preprocessing and select from a range of machine learning models such as Logistic Regression, SVM, KNN, Decision Tree, and Random Forest. The app also offer0s optional hyperparameter tuning and provides evaluation metrics like accuracy, precision, recall, F1-score, confusion matrix, and ROC curve.

This project is open source and licensed under the MIT License, ensuring research transparency with full code and manuscript included. It aims to address the global health crisis of diabetes by enabling early and accurate detection through machine learning.