The Automated Solar Activity Prediction System (ASAP) Update Based on Optimization of a Machine Learning Approach

Abstract

Quite recently, considerable attention has been paid to solar flare prediction because extreme solar eruptions could affect our daily life activities and on different technologies. Therefore, this paper presents a novel method of the development of improved second-generation of the Automated Solar Activity Prediction system (ASAP). The suggested algorithm improves the ASAP system by expanding a period of training vector and generating new machine learning rules to be more successful. Two neural networks are responsible for determining whether the sunspots group will release flare as well as determining if the flare is an M-class or X-class. Several measurement criteria are applied to determine the extent of system performance also all results are provided in this paper. Furthermore, the quadratic score (QR) is used as a metric criterion to compare between the prediction of the proposed algorithm with the Space Weather Prediction Center (SWPC) between 2012 and 2013. The results exhibit that the proposed algorithm outperforms the old ASAP system. Keywords: Solar flares, Machine Learning, Neural network, Space, Prediction, weather.