SHORT-TERM SOLAR FLARE PREDICTION USING MULTIRESOLUTION PREDICTORS
- Harbin Institute of Technology, No. 92 West Da-Zhi Street, Harbin, Heilongjiang Province (China)
- National Astronomical Observatories, 20A Datun Road, Chaoyang District, Beijing (China)
Multiresolution predictors of solar flares are constructed by a wavelet transform and sequential feature extraction method. Three predictors-the maximum horizontal gradient, the length of neutral line, and the number of singular points-are extracted from Solar and Heliospheric Observatory/Michelson Doppler Imager longitudinal magnetograms. A maximal overlap discrete wavelet transform is used to decompose the sequence of predictors into four frequency bands. In each band, four sequential features-the maximum, the mean, the standard deviation, and the root mean square-are extracted. The multiresolution predictors in the low-frequency band reflect trends in the evolution of newly emerging fluxes. The multiresolution predictors in the high-frequency band reflect the changing rates in emerging flux regions. The variation of emerging fluxes is decoupled by wavelet transform in different frequency bands. The information amount of these multiresolution predictors is evaluated by the information gain ratio. It is found that the multiresolution predictors in the lowest and highest frequency bands contain the most information. Based on these predictors, a C4.5 decision tree algorithm is used to build the short-term solar flare prediction model. It is found that the performance of the short-term solar flare prediction model based on the multiresolution predictors is greatly improved.
- OSTI ID:
- 21392363
- Journal Information:
- Astrophysical Journal, Vol. 709, Issue 1; Other Information: DOI: 10.1088/0004-637X/709/1/321; ISSN 0004-637X
- Country of Publication:
- United States
- Language:
- English
Similar Records
A MAGNETOHYDRODYNAMIC MODEL OF THE 2006 DECEMBER 13 ERUPTIVE FLARE
THE NATURE OF FLARE RIBBONS IN CORONAL NULL-POINT TOPOLOGY