Authors: Pouya Hosseinzadeh (Utah State University), Soukaina Filali Boubrahimi (Utah State University), Shah Muhammad Hamdi (Utah State University)

Solar energetic particles (SEPs), which arise from significant solar events, can cause substantial damage to both space-based and terrestrial infrastructure. Predicting these events is challenging due to their rarity and the limited data available. This study improves the prediction of ~30, ~60, and ~100 MeV SEP events by synthetically augmenting the dataset. By employing machine-learning techniques, particularly time series forest (TSF), alongside data augmentation methods like the synthetic minority oversampling technique (SMOTE), we achieved significant improvements in prediction accuracy and F1-score, reaching approximately 90% accuracy for ~100 MeV SEP events.