SUNSPOT 🌞

Sunspots are temporary dark spots on the Sun's photosphere, indicating intense magnetic activity and serving as crucial indicators of solar dynamics and space weather patterns. The dataset used in this project contains monthly mean relative sunspot numbers spanning from 1749 to 1983, meticulously collected by the Swiss Federal Observatory in Zurich until 1960 and subsequently by the Tokyo Astronomical Observatory.

A novel approach to sunspot activity prediction using Temporal Fusion Transformers (TFT) and backtesting strategies.

🎯 Project Aims

1. Develop an accurate time series forecasting model for sunspot activity prediction
2. Implement and evaluate TFT (Temporal Fusion Transformer) for complex temporal patterns
3. Validate predictions through rigorous backtesting
4. Explore the relationship between model confidence and prediction accuracy

🔍 Novel Aspects

• Advanced Architecture: Utilizes state-of-the-art Temporal Fusion Transformer architecture for time series forecasting
• Integrated Backtesting: Combines ML predictions with trading strategy validation
• Quantile Predictions: Provides uncertainty estimates through quantile regression
• Multi-horizon Forecasting: Capable of both short and long-term predictions

📦 Installation

git clone https://github.com/salilp42/SUNSPOT.git
cd SUNSPOT
pip install -r requirements.txt

🚀 Usage

from sunspot_predictor import SunspotPredictor

# Initialize predictor
predictor = SunspotPredictor()

# Load and preprocess data
train, val = predictor.load_data()

# Train model
predictor.create_model()
predictor.train_model()

# Evaluate performance
mape_score, rmse_score = predictor.evaluate_model()

🏗️ Project Structure

SUNSPOT/
├── README.md
├── LICENSE
├── requirements.txt
├── sunspot_predictor.py    # Main implementation
└── sunspot_results/        # Directory for output files

🚧 Work in Progress

This project is actively under development. Current focus areas:

• Optimizing model hyperparameters
• Implementing robust error handling in backtesting
• Adding more sophisticated trading strategies
• Expanding the feature engineering pipeline

🛠 Technical Stack

• Python 3.11+
• PyTorch
• Darts (Time Series Library)
• Backtesting.py
• Pandas & NumPy

📊 Data Source

Dataset Details

• Time Range: 1749-1983
• Frequency: Monthly observations
• Collection: Swiss Federal Observatory, Zurich (1749-1960) and Tokyo Astronomical Observatory (1960-1983)
• Format: Time series of monthly mean relative sunspot numbers
• Source: R datasets package (version 3.6.1)

For more recent sunspot numbers and extended datasets, refer to:

• sunspot.month: Longer series with more recent data
• sunspot.year: Annual aggregated observations

Data Source: R Documentation - Sunspots Dataset

📝 License

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

👤 Author

• salilp42

🤝 Contributing

This is a work in progress and contributions are welcome! Please feel free to submit a Pull Request.