Automatic Short-Term Solar Flare Prediction Using Machine Learning and Sunspot Associations

Abstract

Abstract. In this paper, a machine learning-based system that could provide automated short-term solar flares prediction is presented. This system accepts two sets of inputs: McIntosh classification of sunspot groups and solar cycle data. In order to establish a correlation between solar flares and sunspot groups, the system explores the publicly available solar catalogues from the National Geophysical Data Centre (NGDC) to associate sunspots with their corresponding flares based on their timing and NOAA numbers. The McIntosh classification for every relevant sunspot is extracted and converted to a numerical format that is suitable for machine learning algorithms. Using this system we aim to predict if a certain sunspot class at a certain time is likely to produce a significant flare within six hours time and whether this flare is going to be an X or M flare. Machine learning algorithms such as Cascade-Correlation Neural Networks (CCNN), Support Vector Machines (SVM) and Radial Basis Function Networks (RBFN), are optimised and then compared to determine the learning algorithm that would provide the best prediction performance. It is concluded that SVM provides the best performance for predicting if a McIntosh classified sunspot group is going to flare or not but CCNN is more capable of predicting the class of the flare to erupt. A hybrid system that combines SVM and CCNN is suggested for future use.

Solar Weather Event Modelling and Prediction

“…An automatic short-term solar flare prediction system based on machine learning and sunspot associations has been set up by Qahwaji and Colak (2007). The system can provide a prediction up to six hours in advance by analysing the latest sunspot data, and it confirmed the direct relation between the flare production and certain McIntosh classes like Ekc, Fki and Fkc.…”

“…The efforts made until now to forecast solar flares are often based on different conceptual procedures (see, e.g., Shaw 1989;Miller 1989;Moon et al 2001;Wheatland 2001;Gallagher et al 2002;Jensen et al 2004;Tobiska 2005;Wheatland 2005;Jing et al 2006;Ternullo et al 2006;Barnes et al 2007;Falconer et al 2007;Georgoulis and Rust 2007;Qahwaji and Colak 2007;Colak and Qahwaji 2009).…”

Space Weather: Physics, Effects and Predictability

“…Messerotti et al (2009) have briefly reviewed the forecasting methods. The recently developed Automated Solar Activity Prediction (ASAP) is an upgrade of the prediction system of Qahwaji and Colak (2007) which is based on machine learning and sunspot measurements. ASAP is a hybrid system composed of two neural networks, which provide both the flaring probability of each sunspot group and the relevant flare intensity.…”

Solar flare forecasting using sunspot-groups classification and photospheric magnetic parameters

“…Kasper & Balasubramaniam (2010) found out that the penumbral area, umbral area and irradiance showed promise as possible parameters for predicting solar flares, particularly M-class flares. Qahwaji & Colak (2007) compare the performances of several machine learning algorithm on flare forecasting using classification of sunspot groups and solar cycle data. They found out that Support Vector Machines provide the best performance for predicting whether a classified sunspot group is going to flare.…”

“…To forecasting a flare event is usually converted to classify one sample as a flaring sample or a non-flaring sample. Previously, researchers usually adopt support vector machines, such as Qahwaji & Colak (2007), or neural networks, such as Wang et al (2008). The outputs of support vector machines and neural networks are binary labels indicating flaring or nonflaring.…”