Forecasting SEP Events based on Merged CME Catalogs using Machine Learning

Abstract

The lack of preparation for a Solar Energetic Particle (SEP) event may be catastrophic for astronauts and aircraft passengers alike, along with their electronic equipment. It is widely theorized that SEP events are caused by Coronal Mass Ejections (CMEs), some occurring up to a full day beforehand, accompanied by additional space weather conditions. The only significant models for SEP forecasting are statistically or machine learning-based, often developed on imprecise data. We present an enhanced catalog of CMEs, along with other space weather phenomena, and their relationship with the occurrence of SEP events. Using the enhanced CME catalog, we combine machine learning techniques to create a model that achieves a TSS of 0.829, HSS of 0.712, and F1 Score of 0.714. Further, we analyze the model to determine the relative importance of each input measurement when making SEP occurrence predictions.