期刊文献+

太阳耀斑预报深度学习建模中样本不均衡研究

Study of Sample Imbalance in Deep Learning Modeling of Solar Flare Forecasting
下载PDF
导出
摘要 不同等级耀斑发生的频次存在数量级上的差别,使基于常规卷积神经网络的耀斑预报模型通常难以捕捉M和X类耀斑先兆特征,导致高等级耀斑预报精度低的问题.本文对于这种耀斑预报中的长尾分布问题,通过控制变量法讨论不同深度长尾学习方法对于耀斑预报精度提升.尝试从训练集优化、损失函数优化、网络权重优化等角度改进模型对于M和X类耀斑的预报性能.在SDO/HMI太阳磁图预报未来24 h耀斑的实验中,相比于常规方法训练的基准模型,改进模型在M和X类耀斑预报的精确率分别有了53.10%和38.50%的提升,同时在召回率上有64%和52%的提升.表明在耀斑预报问题中,数据长尾分布的处理至关重要,验证了深度长尾学习方法的有效性.这种提升尾部类预报精确率的方法不仅可以应用于耀斑预报领域,还可以迁移到其他具有长尾分布现象的空间天气典型事件的预报分析中. Solar flares,as violent eruptions occurring in the lower atmosphere of the Sun,exert signifi-cant impacts on human activities.Researchers globally have developed multiple prediction models for so-lar flares,employing empirical,physical,statistical,and other methodologies.There is an order of magni-tude difference in the occurrence of different classes of flares.This makes it difficult for traditional con-volutional neural network-based flare prediction models to capture M,X class flare features,which leads to the problem of low precision of high level flare prediction.With the breakthrough of deep learning technology in recent years,it has shown strong potential in modelling and prediction of complex prob-lems and a number of works have begun to try to use deep learning methods to construct flare predic-tion models.In this paper,different deep long-tail learning methods are discussed by us to improve the precision of flare forecasting by controlling the variables for the long-tail distribution phenomenon in flare forecasting.The forecast performance of the model for M and X flares is tried to be improved from the perspectives of training set optimization,loss function optimization and network weight optimization.The experiments on SDO/HMI solar magnetogram data show that the precision of M,X class flare pre-diction is significantly improved by 53.10%and 38.50%,respectively,and the recall is increased by 64%and 52%compared with the baseline model trained by conventional methods.It shows that the treat-ment of the long-tailed distribution of data is crucial in the flare forecasting problem,and verifies the ef-fectiveness of the deep long-tailed learning method.This method of improving the precision of tail class forecasts can be applied not only to the field of flare forecasting,but also can be transferred to the analy-sis of forecasting other typical events of space weather with long-tailed distribution phenomenon.
作者 周俊 佟继周 李云龙 方少峰 ZHOU Jun;TONG Jizhou;LI Yunlong;FANG Shaofeng(National Space Science Center,Chinese Academy of Sciences,Beijing 100190;University of Chinese Academy of Sciences,Beijing 100049;National Space Science Data Center,Beijing 101407)
出处 《空间科学学报》 CAS CSCD 北大核心 2024年第2期241-250,共10页 Chinese Journal of Space Science
基金 国家重点研发计划项目(2022YFF0711400) 中国科学院网信专项(CAS-WX2022SF-0103)共同资助。
关键词 耀斑预报 长尾分布 残差神经网络 Flare prediction Long-tailed distribution Residual neural network
  • 相关文献

参考文献4

二级参考文献28

  • 1Abramenko, V. I., Yurchyshyn, V. B., Wang, H., Spirock, I". J., & Goode, P. R. 2003, ApJ, 597, 1135.
  • 2Barnes, G., Leka, K. D., Schumer, E. A., & Della-Rose, D. J. 2007, Space Weather, 5, 9002.
  • 3Boser, B., Guyon, I., & Vapnik, V. 1992, A training algorithm for optimal margin classifiers, in the Fifth Annual Workshop on Computational Learning Theory, 144.
  • 4Chang, C.-C., & Lin, C.-J. 2001, LIBSVM: a library for support vector machines, software available at http://www, csie.ntu, edu. tw/ cjlin/libsvm.
  • 5Chen, W.-Z., Liu, C., Song, H., et al. 2007, ChJAA (Chin. J. Astron. Astrophys.), 7, 733.
  • 6Cortes, C., & Vapnik, V. N. 1995, Support-vector network, Machine Learning, 20, 273.
  • 7Dauphin, C., Vilmer, N., & Anastasiadis, A. 2007, A&A, 468, 273.
  • 8Falconer, D. A., Moore, R. L., & Gary, G. A. 2003, J. Geophys. Res. (Space Physics), 108, 1380.
  • 9Fan, R. E., Chert, P. H., & Lin, C. J. 2005, Journal of Machine Learning Research, 6, 1889.
  • 10Gallagher, P. T., Moon, Y., & Wang, H. 2002, Sol. Phys., 209, 171.

共引文献9

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部