Title and Abstract

Title: ML Efforts on Understanding and Predicting Solar Flares

Abstract: In this talk, I present our machine learning efforts, which show great promise towards early predictions of solar flare events. First, we develop a data pre-processing pipeline that is built to extract useful data from multiple sources -- GOES, SDO/HMI, SDO/AIA -- to prepare inputs for general-purpose machine learning algorithms. Second, in our strong/weak flare classification model, case studies show a significant increase in the prediction score around 20 hours before strong solar flare events, which implies that early precursors appear at least 20 hours prior to the peak of a flare event. Third, we innovate off-the-shelf machine learning algorithms and propose a mixed LSTM regression model to predict solar flare intensity jointly with flaring/non-flaring indicator. Last, I will present our results on interpretable ML methods that have direct physical meanings. Ongoing and future work will be briefly discussed if time permits.