Flame-Watch is an interactive web application for wildfire prediction, powered by a custom-trained machine learning model. The application is built using Django and provides an easy-to-use interface for users to predict wildfire risks based on provided data. The ML models and training scripts are included in the Model/ directory, allowing users to explore, retrain, or understand the prediction logic.

• Wildfire risk prediction using a custom ML model
• Interactive web interface built with Django
• Reference datasets and model training scripts included
• Easy local setup and deployment

Application/
    db.sqlite3
    manage.py
    requirements.txt
    App/
        __init__.py
        admin.py
        apps.py
        best_model_tuned.pkl
        best_model.pkl
        forms.py
        Ml.py
        models.py
        tests.py
        urls.py
        views.py
        migrations/
        templates/
    Flamewatch/
        __init__.py
        asgi.py
        settings.py
        urls.py
        wsgi.py
    public/
    static/
    tmp/

Model/
    best_model_tuned.pkl
    best_model.pkl
    merging.py
    nasa_data_nonf.py
    nasa_data.py
    testing_m.py
    testing_usingdat.py
    training.py
    Training Dataset/
        nasa_merged_wildfire_data_1.csv
        nasa_non_fire_data_tst.csv
        ...

1. Install Python

   • Recommended: Python 3.11 or newer

2. Clone the repository

   git clone https://github.com/DvTanviR/Flame-Watch.git
   cd Flame-Watch/Application

3. (Optional but recommended) Create a virtual environment

   python -m venv venv
   venv\Scripts\activate  # Windows
   # source venv/bin/activate  # Linux/macOS

4. Install dependencies

   pip install -r requirements.txt

5. Apply migrations

   python manage.py migrate

6. Run the development server

   python manage.py runserver

   • Access the app at http://127.0.0.1:8000/

1. Clone the repository

   git clone https://github.com/DvTanviR/Flame-Watch.git
   cd Flame-Watch/Model

2. Create a Python virtual environment (recommended)

   python -m venv venv
   venv\Scripts\activate  # Windows
   # source venv/bin/activate  # Linux/macOS

3. Install dependencies

   pip install -r requirements.txt

4. Train the model (optional)

   python Model/training.py

5. Run tests or make predictions

   python Model/testing_usingdat.py

You can deploy the Django application using any standard method (Heroku, PythonAnywhere, VPS, etc.). No special steps or API keys are required. All necessary files are packaged in the Application/ folder.

• Pre-trained models: best_model.pkl, best_model_tuned.pkl (see Model/)
• Training scripts and datasets: Provided in Model/ for reference and experimentation

These metrics indicate the model's strong capability in correctly predicting wildfire occurrences, balancing false positives and false negatives effectively.

1. Data ingestion: Satellite and weather data from NASA APIs are preprocessed and merged.
2. Feature extraction: Important variables (temperature, humidity, NDVI, wind speed, etc.) are selected.
3. Model training: A Random Forest classifier learns patterns distinguishing wildfire vs. non-wildfire conditions.
4. Prediction: The trained model evaluates new input data and predicts the risk of wildfire.
5. Output: Risk scores can be visualized or integrated into alerting systems.

• Wildfire risk prediction: Predicts fire occurrence based on climate variables like temperature, humidity, wind speed, and satellite thermal hotspot data.
• NASA data integration: Uses publicly available NASA FIRMS and POWER datasets.
• Random Forest classifier: Proven machine learning algorithm providing high accuracy and interpretability.
• Balanced dataset training: Trained with both fire and no-fire data to reduce false positives.
• Performance metrics: Achieves over 90% accuracy, precision, and recall in validation tests.
• MIT licensed: Free for commercial and non-commercial use.

To understand or retrain the ML model, explore the scripts and datasets in the Model/ directory. The Django app in Application/ is ready to use and can be customized as needed.

For any questions or contributions, feel free to open an issue or pull request.