Using X-Ray Flux Time Series for Solar Explosion Forecasting

Among the natural phenomena that happen in the Sun, one that has direct impact on Earth are the solar explosions. These explosions emit an X-ray flux that can be detected and used as an indicative of new explosions. In this work, we evaluated whether X-ray flux time series are suitable as the base dataset for solar flare forecasting. To do so, we applied a multilayer perceptron (MLP) neural network and performed a series of experiments to identify its best parameters and its performance for different forecast horizons. The experiments indicated that X-ray flux time series can be used to forecast solar flares and that MLPs are capable of properly forecasting future values of the time series, even though the average errors increase with wider forecast horizons. Besides, we also observed that qualitative analyses are essential for solar explosion forecasts, and should also be made together with traditional quantitative analyses of the results.