Diabetes-Prediction-ML: A Complete Multi-End Diabetes Prediction System

This project is a machine learning-based diabetes risk assessment system with three interaction modes: desktop application, Web interface, and RESTful API. It uses SMOTE to handle data imbalance issues, and the decision tree model achieves an accuracy of 78.2%. Key basic information: Maintained by Ahmed Elsaadany (@A7medElsaadany), hosted on GitHub (original link: https://github.com/A7medElsaadany/Diabetes-Prediction-ML), released on June 17, 2026.

Diabetes is one of the most common chronic diseases globally, and early detection is crucial for improving patient prognosis. This project provides a complete machine learning solution for diabetes risk assessment. Origin details: Maintained by Ahmed Elsaadany (@A7medElsaadany), source platform is GitHub, original title is Diabetes-Prediction-ML, released on June 17, 2026.

The project adopts a modular design, decoupling the machine learning core from multiple user interfaces. Three interaction modes:

1. Desktop application (Tkinter GUI): Built with Python's Tkinter library, has 8 input fields covering key medical indicators like blood glucose and blood pressure, supports real-time result return and dark theme.
2. Web interface (Flask): Responsive Web app using Flask 3.0.0, optimized for mobile devices, with features like real-time prediction and animated confidence bar.
3. RESTful API: Provides POST /predict endpoint with JSON support and CORS configuration, enabling integration with third-party systems like hospital information systems and health management apps.

Data Preprocessing: Uses SMOTE to handle class imbalance (health samples are more than disease samples) and StandardScaler for feature standardization. Model Selection: Compared 7 algorithms (Logistic Regression, SVM, KNN, Decision Tree, Naive Bayes, Random Forest, Gradient Boosting), and selected Decision Tree for balanced performance. Performance Metrics:

Metric	Score
Accuracy	78.2%
Precision	76.5%
Recall	79.1%
F1 Score	77.8%

Toolchain:

Component	Tech Stack
Backend API	Flask 3.0.0
Machine Learning	Scikit-learn 1.8.0
Data Processing	Pandas, NumPy
Data Balance	Imbalanced-learn (SMOTE)
Visualization	Matplotlib
Desktop GUI	Tkinter
Web Frontend	HTML5, CSS3, JavaScript
Model Serialization	Joblib
Explainability: Includes feature importance analysis to help understand prediction factors, which is critical for medical AI applications (building trust with doctors and patients, supporting medical research).

Application Scenarios:

1. Community health screening (identify high-risk groups)
2. Medical institution auxiliary diagnosis (provide second opinion)
3. Health management app integration (via API)
4. Medical education (typical case of ML in healthcare) Deployment Notes: The prediction results are for reference only and cannot replace professional medical diagnosis (model trained on historical data, may not cover all individual differences and special cases).

The Diabetes-Prediction-ML project demonstrates a complete ML application development process from data preprocessing and model training to multi-platform deployment, providing valuable reference for medical AI engineering. Highlights: Multi-end delivery (same core logic for different user groups). Future improvements: Introduce deep learning models, add more feature dimensions, support online model updates, and add model drift detection mechanisms.