Skip to content

README.md

Latest commit

 

History

History
327 lines (242 loc) · 10.4 KB

README.md

File metadata and controls

327 lines (242 loc) · 10.4 KB

🏥 Health-Insight — ML-Powered Disease Prediction

Python Flask Scikit-learn Pandas NumPy License

An end-to-end ML-powered web application that predicts risk for multiple diseases in real time.

Live Demo · Report Bug · Request Feature

📌 Table of Contents

📖 About the Project

Health-Insight is a full-stack web application that integrates multiple machine learning models with a Flask backend to deliver real-time disease risk predictions directly in the browser.

Designed with production-style considerations in mind, this project addresses challenges like feature consistency between training and inference, model serialization, dynamic form generation, and robust error handling — making it more than just a demo, but a template for real-world ML deployment.

⚠️ Disclaimer: This application is intended for educational and research purposes only. It is not a substitute for professional medical advice, diagnosis, or treatment. Always consult a qualified healthcare provider.

🦠 Supported Diseases

Disease Model File Dataset
🩸 Diabetes diabetes.pkl diabetes.csv
❤️ Heart Disease heart.pkl heart.csv
🫘 Kidney Disease kidney.pkl kidney.csv
🫀 Liver Disease liver.pkl liver.csv
🔬 Cancer cancer.pkl cancer.csv

✨ Features

  • 🤖 Individual ML model per disease — each condition has its own dedicated RandomForestClassifier
  • 📋 Dynamic input forms — automatically generated based on disease-specific feature sets
  • ⚡ Real-time predictions — instant inference via Flask REST backend
  • 💾 Model persistence — serialized and loaded using Pickle for fast startup
  • ✅ Input validation — client and server-side checks before inference
  • 🔡 Categorical encoding — handles mixed data types in production
  • 🧱 Modular architecture — clean separation of training, inference, and UI layers
  • 📱 Responsive UI — works across desktop and mobile browsers

🛠 Tech Stack

Layer Technology
Backend Python 3.x, Flask
ML Framework Scikit-learn (RandomForestClassifier)
Data Handling Pandas, NumPy
Serialization Pickle
Frontend HTML5, CSS3, JavaScript
Templating Jinja2

⚙️ How It Works

User selects a disease
        │
        ▼
Dynamic form rendered with disease-specific input fields
        │
        ▼
User submits health parameters
        │
        ▼
Flask validates & encodes inputs
        │
        ▼
Correct .pkl model loaded for the selected disease
        │
        ▼
RandomForestClassifier runs inference
        │
        ▼
Prediction result displayed on result.html
  1. Training Phase — Each disease has a standalone training script (training/<disease>.py) that preprocesses the dataset, trains a RandomForestClassifier, and saves the model as a .pkl file.
  2. Inference Phase — When a user submits a form, Flask loads the corresponding .pkl model, applies the same preprocessing pipeline, and returns the prediction.
  3. Feature Consistency — Feature names and encoding schemes are kept consistent between training and inference to prevent silent prediction errors.

🚀 Getting Started

Prerequisites

Ensure the following are installed on your system:

Installation

  1. Clone the repository:
git clone https://github.com/your-username/Health-Insight.git
cd Health-Insight
  1. Create and activate a virtual environment (recommended):
python -m venv venv
source venv/bin/activate        # macOS/Linux
venv\Scripts\activate           # Windows
  1. Install all dependencies:
pip install -r requirements.txt
  1. (Optional) Retrain the models:
python training/diabetes.py
python training/heart.py
python training/kidney.py
python training/liver.py
python training/cancer.py

Pre-trained .pkl files are included in the models/ directory so retraining is optional.

💻 Usage

  1. Start the Flask development server:
python app.py
  1. Open your browser and visit:
http://127.0.0.1:5000/
  1. Select a disease, fill in the health parameters, and click Predict to receive your risk assessment instantly.

📁 Project Structure

Health-Insight/
│
├── app.py                    # Main Flask app — routes & inference logic
├── requirements.txt          # All Python dependencies
│
├── models/                   # Serialized trained ML models
│   ├── diabetes.pkl
│   ├── heart.pkl
│   ├── kidney.pkl
│   ├── liver.pkl
│   └── cancer.pkl
│
├── training/                 # Standalone training scripts per disease
│   ├── diabetes.py
│   ├── heart.py
│   ├── kidney.py
│   ├── liver.py
│   └── cancer.py
│
├── datasets/                 # Raw CSV datasets used for training
│   ├── diabetes.csv
│   ├── heart.csv
│   ├── kidney.csv
│   ├── liver.csv
│   └── cancer.csv
│
├── templates/                # Jinja2 HTML templates
│   ├── index.html            # Landing page — disease selector
│   ├── form.html             # Dynamic input form
│   └── result.html           # Prediction result display
│
├── static/                   # Static assets
│   ├── css/                  # Stylesheets
│   └── js/                   # JavaScript files
│
└── README.md

🔌 API Endpoints

Method Endpoint Description
GET / Home page — disease selection
GET /predict/<disease> Load input form for selected disease
POST /predict/<disease> Submit form and return prediction result

Example POST body for diabetes prediction:

{
  "pregnancies": 2,
  "glucose": 138,
  "blood_pressure": 62,
  "skin_thickness": 35,
  "insulin": 0,
  "bmi": 33.6,
  "diabetes_pedigree": 0.627,
  "age": 47
}

📊 Model Performance

Results from training on the provided datasets. Metrics may vary with different train/test splits.

Disease Algorithm Accuracy
Diabetes RandomForestClassifier ~76–80%
Heart Disease RandomForestClassifier ~82–86%
Kidney Disease RandomForestClassifier ~96–99%
Liver Disease RandomForestClassifier ~72–76%
Cancer RandomForestClassifier ~94–97%

🧠 Key Learnings

  • Feature consistency — Ensuring training feature names/order exactly match inference inputs to prevent silent errors
  • Categorical encoding in production — Handling label encoding and one-hot encoding at inference time without refitting
  • Real-world ML deployment — Debugging shape mismatches, missing values, and dtype inconsistencies
  • Modular backend design — Separating training logic from inference for clean, maintainable code
  • Flask routing patterns — Building dynamic, parameterized routes for multi-model applications

🔮 Future Improvements

  • REST API endpoints with JSON responses for mobile/external integration
  • User authentication and prediction history dashboard
  • Model monitoring, drift detection & automated retraining pipeline
  • Dockerized deployment with docker-compose
  • SHAP-based explainability — show which features drove the prediction
  • Confidence scores alongside binary predictions
  • CI/CD pipeline with GitHub Actions

🤝 Contributing

Contributions are welcome and appreciated!

  1. Fork the repository
  2. Create a feature branch: git checkout -b feature/your-feature
  3. Commit your changes: git commit -m "Add: your feature description"
  4. Push to your branch: git push origin feature/your-feature
  5. Open a Pull Request

Please follow PEP 8 coding standards and include docstrings for any new functions.

📄 License

Distributed under the MIT License. See LICENSE for more information.

📬 Contact

Your Name — ganesh1a0576@gmail.com

GitHub: Ganesh-a0576

Project Link: https://github.com/your-username/Health-Insight

⭐ If this project helped you, please consider giving it a star — it means a lot!


Made with ❤️ and Python