Using Support Vector Machine (SVM) and Ionospheric Total Electron Content (TEC) Data for Solar Flare Predictions

Predicting where and when space weather events such as solar flares X-rays bursts are likely to occur in a specific area of interest constitutes significant challenge research. Space scientists are, therefore, gradually exploring multivariate data analysis techniques from the fields mining or machine learning order approximate future occurrences past distribution patterns. As emit extreme ultraviolet X-ray radiation, which leads ionization effect different layers ionosphere, most recent works related flare predictions using (ML) techniques, focused on time series predictions. Here, we suggest support vector for classifying subdaily diurnal total electron content (TEC) spatial changes prior events, assess possibility predicting B, C, M, X-class events. This is done opposed TEC ML techniques. The estimated up three days before each tested class along with skill scores precision, recall, Heidke score (HSS), accuracy, true statistics. results indicate that suggested approach has ability predict X M-class 24 h their occurrence 91% 76% HSS scores, respectively, improves over works. However, small-size C B-class flares, does not succeed producing similar promising results.