SEP Occurrence Forecasting

This example is adapted from Aminalragia-Giamini et al. (2021), Solar Energetic Particle Event occurrence prediction using Solar Flare Soft X-ray measurements and Machine Learning.

The saved data contain 49 numerical features, binary labels, and an existing train/test assignment. The notebooks can therefore demonstrate sample-level classification, class imbalance, validation, and SHAP mechanics, but they cannot establish event-aware performance or physical feature attribution. For a more detailed discussion contact the authors.

• Native PyTorch neural workflow

• Keras 3 neural alternative using the Torch backend

• XGBoost, sample-level validation, and SHAP

Reference: S. Aminalragia-Giamini et al. (2021), Journal of Space Weather and Space Climate, 11, 59.