<![CDATA[Solar flares (SFs) are the most powerful bursts of energy in the solar system that often have a bad effect on space weather. Until now, the cause of its appearance is not known for sure. Nevertheless, SFs are known to have magnetic properties attached to them. Therefore, understanding the configuration of the magnetic field on the sun plays an important role in SFs prediction efforts. Using SFs flux data recorded by X-ray Sensors on the Geostationary Operational Environmental Satellite (GOES) which is mapped with 13 parameters of the magnetic vector data of the solar photosphere layer recorded by the Helioseismic and Magnetic Imager (HMI) at the Solar Dynamic Observatory (SDO) and the Machine Learning (ML) Random Ferns (RFe) algorithm, This study tries to predict the emergence of multiclass SFs (B, C, M, and X) along with binary SFs (BC and MX). This study uses data from May 1, 2010 to May 10, 2020, with a total of 30 classes X, 443 classes M, 1032 classes C, 751 classes B, 473 classes MX, and 1783 classes BC. This study also applies the oversampling method to handle the imbalanced nature of the data on SFs data. Overall, it can be seen that predicting the occurrence of SFs using RFe is a valid effort. The highest average scores achieved by this study for sensitivity/recall, precision, and True Skill Statistics (TSS) in multiclass SFs were 74.4%, 50.3%, and 58.7%, respectively; and in binary SFs are 87.7%, 77.7%, and 72.8%.]]>
Copyrights © 2023
