hepatitis-c-predictor 0.4.2

Interactive ML application for Hepatitis C classification using PyTorch with Streamlit interface

Hepatitis C Predictor

Developing a Deep Learning model with PyTorch to identify patterns and predict the presence of Hepatitis C from patient data, paving the way for faster, more accurate diagnostics.

Features

• 📊 Interactive Data Exploration: Visualize and explore the Hepatitis C dataset
• 🚀 Model Training Interface: Train models with custom hyperparameters
• 📈 Model Evaluation: Comprehensive performance metrics and visualizations
• 🤖 Deep Learning: PyTorch-based neural network with residual connections
• 📦 Auto-download: Dataset downloads automatically if not present

Dataset

The dataset contains laboratory values from blood donors and Hepatitis C patients:

• Source: UCI Machine Learning Repository / Kaggle
• Size: 615 samples
• Features: 12 laboratory measurements + age and sex
• Target: Binary classification (Healthy vs Hepatitis C)
• Auto-download: The app will automatically download the dataset if not present

Model

• Architecture: Deep Neural Network with Residual Connections
  • Input Layer: 12 features
  • Hidden Layers: [128, 64, 32] neurons
  • Residual Blocks: 2 per hidden layer
  • Output Layer: 2 classes (Binary classification)
• Framework: PyTorch 2.8+
• Regularization: Layer Normalization + Dropout (0.3)
• Expected Accuracy: ~97.5% on validation set

🚀 Deployment

Requirements for Deployment

• Python 3.10
• All dependencies listed in requirements.txt
• Dataset will be downloaded automatically on first run

License

This project is licensed under the MIT License - see the LICENSE file for details.

Important Note

⚠️ This model is for educational purposes only. Do not use for actual medical diagnosis. Always consult healthcare professionals.

To run it locally:

uvx --from hepatitis-c-predictor --index-url https://test.pypi.org/simple/ --extra-index-url https://pypi.org/simple/ hepatitis-c-demo

Docs

You can check the modules docs in the docs folder or directly from the deployed version on GH pages here: https://ninjalice.github.io/HEPATITIS_C_MODEL/src.html

Getting Started

1. Clone the repository:

   git clone https://github.com/Ninjalice/HEPATITIS_C_MODEL.git
   cd HEPATITIS_C_MODEL
   

2. Install dependencies:

   pip install -r requirements.txt
   

   Or using uv:

   uv sync --frozen
   

3. Run the interactive dashboard:

   streamlit run app.py
   

   The app will automatically download the dataset if not present.

Option 2: Jupyter Notebooks

Follow the notebooks in order:

1. 01-data-exploration.ipynb - Explore the dataset
2. 02-data-preprocessing.ipynb - Clean and prepare data
3. 03-model-training.ipynb - Train the neural network
4. 04-model-prediction.ipynb - Make predictions on new data (WIP)

Manual Download (Optional)

If auto-download fails, you can manually download from:

1. Kaggle: https://www.kaggle.com/datasets/fedesoriano/hepatitis-c-dataset
2. Place the file in data/raw/hepatitis_data.csv

Project Organization

├── data/
│   ├── raw/              <- The original, immutable data dump
│   └── processed/        <- The final, canonical data sets for modeling
│
├── models/               <- Trained and serialized models
│
├── notebooks/            <- Jupyter notebooks for analysis and modeling
│   ├── 01-data-exploration.ipynb
│   ├── 02-data-preprocessing.ipynb
│   ├── 03-model-training.ipynb
│   └── 04-model-prediction.ipynb
│
├── reports/              <- Generated analysis as HTML, PDF, LaTeX, etc.
│   └── figures/          <- Generated graphics and figures
│
├── src/                  <- Source code for use in this project
│   ├── __init__.py
│   ├── data.py           <- Scripts to download or generate data
│   ├── train.py          <- Scripts to train models
│   ├── models.py         <- Scripts to train models and make predictions
│   └── visualization.py  <- Scripts to create exploratory visualizations
│
├── requirements.txt      <- The requirements file for reproducing the environment
└── README.md             <- The top-level README for developers