Hostname: page-component-7b7567c8c-z4zkx Total loading time: 0 Render date: 2026-11-17T14:27:52.422Z Has data issue: false hasContentIssue false

Solar Flare Prediction Using Machine Learning with Multiwavelength Observations

Published online by Cambridge University Press:  24 July 2018

Naoto Nishizuka
Affiliation:
National Institute of Information and Communications Technology, 4-2-1 Nukui-Kitamachi, Koganei, Tokyo, Japan 184-8795 email: nishizuka.naoto@nict.go.jp
Komei Sugiura
Affiliation:
National Institute of Information and Communications Technology, 4-2-1 Nukui-Kitamachi, Koganei, Tokyo, Japan 184-8795 email: nishizuka.naoto@nict.go.jp
Yuki Kubo
Affiliation:
National Institute of Information and Communications Technology, 4-2-1 Nukui-Kitamachi, Koganei, Tokyo, Japan 184-8795 email: nishizuka.naoto@nict.go.jp
Mitsue Den
Affiliation:
National Institute of Information and Communications Technology, 4-2-1 Nukui-Kitamachi, Koganei, Tokyo, Japan 184-8795 email: nishizuka.naoto@nict.go.jp
Shin-ichi Watari
Affiliation:
National Institute of Information and Communications Technology, 4-2-1 Nukui-Kitamachi, Koganei, Tokyo, Japan 184-8795 email: nishizuka.naoto@nict.go.jp
Mamoru Ishii
Affiliation:
National Institute of Information and Communications Technology, 4-2-1 Nukui-Kitamachi, Koganei, Tokyo, Japan 184-8795 email: nishizuka.naoto@nict.go.jp
Rights & Permissions [Opens in a new window]

Abstract

Core share and HTML view are not available for this content. However, as you have access to this content, a full PDF is available via the 'Save PDF' action button.

We developed a flare prediction model based on the supervised machine learning of solar observation data for 2010-2015. We used vector magnetograms, lower chromospheric brightening, and soft-X-ray data taken by Solar Dynamics Observatory and Geostationary Operational Environmental Satellite. We detected active regions and extracted 60 solar features such as magnetic neutral lines, current helicity, chromospheric brightening, and flare history. We fully shuffled the database and randomly divided it into two for training and testing. To predict the maximum size of flares occurring in the following 24 hours, we used three machine-learning algorithms independently: the support vector machine, the k nearest neighbors (kNN), and the extremely randomized trees. We achieved a skill score (TSS) of greater than 0.9 for kNN. Furthermore, we compared the prediction results in a more operational setting by shuffling and dividing the database with a week unit. It was found that the prediction score depends on the way the database is prepared.