Title :
A new solar flare prediction model based on large-scale image management
Author :
Yu, Daren ; Zhang, Xiaopeng ; Liu, Jinfu ; Wang, Qiang
Author_Institution :
Sch. of Astronaut., Harbin Inst. of Technol., Harbin, China
Abstract :
Solar flare is one of the most powerful solar activities, which plays a very important role in daily life and space weather, so it is meaningful to predict solar flare accurately. In this paper, a new solar flare prediction model based on large-scale image management `LM-Sphere´ has been proposed, which is built by large-margin, image indexing and hypersphere theory. Experimental results show that: 1) LM-Sphere has retained the distribution of original image data which reduced the accuracy loss of sampling in usual large-scale data modeling; 2) The application of large-scale image management method has solved the local minimum point´s problem in K-nearest neighbor (KNN) algorithm which improved the accuracy of flare prediction. All in all, LM-sphere is a powerful model in solar flare prediction.
Keywords :
astronomical image processing; indexing; learning (artificial intelligence); solar flares; LM-sphere; hypersphere theory; image indexing; k-nearest neighbor algorithm; large-scale data modeling; large-scale image management method; solar flare prediction model; space weather; Educational institutions; Flare prediction; Hypersphere; Large margin; Large-scale image indexing;
Conference_Titel :
Software Engineering and Service Science (ICSESS), 2012 IEEE 3rd International Conference on
Conference_Location :
Beijing
Print_ISBN :
978-1-4673-2007-8
DOI :
10.1109/ICSESS.2012.6269477