BCG Severity Prediction Web App - Setup Guide

1. Files Structure

Make sure your project directory has the following structure:

bcg_severity_prediction/
├── app.py             # Flask application
├── prediction.py      # Prediction module
├── templates/         # HTML templates
│   ├── index.html     # Input form
│   ├── result.html    # Prediction results
│   └── error.html     # Error handling
└── models/            # Model files
    └── final_improved_bcg_severity_model.pkl  # Your trained model

Application Screenshots

Input Form

Prediction Result

2. Installation Steps

1. Make sure you have Python 3.8+ installed
2. Install required packages:

   pip install flask pandas numpy scikit-learn lightgbm xgboost imbalanced-learn
   

3. Place your trained model file in the models directory
   • If you've already trained the model using the provided training script, it should be available as final_improved_bcg_severity_model.pkl
   • If not, you can use a simple rule-based fallback in the prediction.py file

3. Running the Application

1. Open a terminal/command prompt
2. Navigate to your project directory:

   cd path/to/bcg_severity_prediction
   

3. Run the Flask application:

   python app.py
   

4. Open your web browser and go to:

   http://127.0.0.1:5000
   

4. Troubleshooting

If you encounter any issues:

1. Model file not found error:

   • The prediction module has a built-in fallback that will provide rule-based predictions if the model file isn't found
   • Make sure your model file is in the correct location (models/final_improved_bcg_severity_model.pkl)

2. Scikit-learn version compatibility issues:

   • If you see errors related to "incompatible dtype" or similar, it means there's a version mismatch between the scikit-learn used to train the model and the version being used now
   • The prediction module has error handling that will fall back to rule-based predictions

3. Missing required fields:

   • The prediction module automatically handles missing fields by using default values

4. Port already in use:

   • Change the port in app.py by modifying the last line to:

     app.run(debug=True, port=5001)

   • Then access the app at http://127.0.0.1:5001

5. Making Predictions

The web app should now provide different predictions based on the input data, with the following key fields having the most impact:

• Age/Years
• Adverse event by location (Local, Systemic, Mixed)
• Route of administration
• System Organ class affected
• Time to onset

Try entering different values to see how they affect the prediction.