Machine Learning Techniques for Predicting the Geoeffectiveness of Coronal Mass Ejections

Machine Learning techniques are a powerful tool for addressing various types of issues. This presentation aims to demonstrate how they have been applied to predict the geoeffectiveness of Coronal Mass Ejections (CMEs). Our research focused on experimenting with various binary classification models on white-light coronagraph data sets of close-to-Sun CMEs. In this presentation, algorithms such as logistic regression, k-nearest neighbors, artificial neural networks, and ensemble models will be discussed, together with their performances. Additionally, the challenges encountered during the research, such as the extreme imbalance between the number of geoeffective and ineffective events in our data set, as well as insights into the approaches taken to overcome them, will be addressed.