Predicting Solar Flares with Machine Learning: Investigating Solar Cycle Dependence
Abstract
A deep learning network, long short-term memory (LSTM), is used to predict whether an active region (AR) will produce a flare of class Γ in the next 24 hr. We consider Γ to be ≥M (strong flare), ≥C (medium flare), and ≥A (any flare) class. The essence of using LSTM, which is a recurrent neural network, is its ability to capture temporal information on the data samples. The input features are time sequences of 20 magnetic parameters from the space weather Helioseismic and Magnetic Imager AR patches. We analyze ARs from 2010 June to 2018 December and their associated flares identified in the Geostationary Operational Environmental Satellite X-ray flare catalogs. Our results produce skill scores consistent with recently published results using LSTMs and are better than the previous results using a single time input. The skill scores from the model show statistically significant variation when different years of data are chosen for training and testing. In particular, 2015-2018 have better true skill statistic and Heidke skill scores for predicting ≥C medium flares than 2011-2014, when the difference in flare occurrence rates is properly taken into account.
- Publication:
-
The Astrophysical Journal
- Pub Date:
- May 2020
- DOI:
- 10.3847/1538-4357/ab89ac
- arXiv:
- arXiv:1912.00502
- Bibcode:
- 2020ApJ...895....3W
- Keywords:
-
- Solar flares;
- Solar activity;
- 1496;
- 1475;
- Astrophysics - Solar and Stellar Astrophysics
- E-Print:
- doi:10.3847/1538-4357/ab89ac