Comparison of the Efficiency of Machine Learning Methods in Studying the Importance of Input Features in the Problem of Forecasting the Dst Geomagnetic Index
- Authors: Vladimirov R.D.1, Shirokiy V.R.1, Myagkova I.N.1, Barinov O.G.1, Dolenko S.A.1
-
Affiliations:
- Skobeltsyn Institute of Nuclear Physics, Lomonosov Moscow State University
- Issue: Vol 63, No 2 (2023)
- Pages: 190-201
- Section: Articles
- URL: https://ruspoj.com/0016-7940/article/view/651026
- DOI: https://doi.org/10.31857/S0016794022100224
- EDN: https://elibrary.ru/DLYMRJ
- ID: 651026
Cite item
Abstract
One of the promising approaches to predicting the values of geomagnetic indices is the use of
machine learning methods. However, for the effective use of such methods, it is necessary to select essential
input features of the problem in order to reduce its input dimension. In this paper, we consider an algorithm
for obtaining the most efficient forecasting model based on lowering the input data dimension by gradually
discarding input features based on the following machine learning methods: linear regression, gradient boosting,
and a multilayer perceptron artificial neural network. The effectiveness of the listed methods is compared;
the directions of further development of this work are considered
About the authors
R. D. Vladimirov
Skobeltsyn Institute of Nuclear Physics, Lomonosov Moscow State University
Email: vladimirov.rd16@physics.msu.ru
Moscow, 119991 Russia
V. R. Shirokiy
Skobeltsyn Institute of Nuclear Physics, Lomonosov Moscow State University
Email: shiroky@srd.sinp.msu.ru
Moscow, 119991 Russia
I. N. Myagkova
Skobeltsyn Institute of Nuclear Physics, Lomonosov Moscow State University
Email: irina@srd.sinp.msu.ru
Moscow, 119991 Russia
O. G. Barinov
Skobeltsyn Institute of Nuclear Physics, Lomonosov Moscow State University
Email: obar@sinp.msu.ru
Moscow, 119991 Russia
S. A. Dolenko
Skobeltsyn Institute of Nuclear Physics, Lomonosov Moscow State University
Author for correspondence.
Email: dolenko@srd.sinp.msu.ru
Moscow, 119991 Russia
References
- − Белов А.В., Виллорези Дж., Дорман Л.И. и др. Влияние космической среды на функционирование искусственных спутников Земли // Геомагнетизм и аэрономия. Т. 44. № 4. С. 502‒510. 2004.
- – Ермолаев Ю.И., Ермолаев М.Ю. Солнечные и межпланетные источники геомагнитных бурь: Аспекты космической погоды // ГЕОФИЗИЧЕСКИЕ ПРОЦЕССЫ И БИОСФЕРА. Т. 8. № 1. С. 5‒35. 2009.
- – Ефиторов А.О., Мягкова И.Н., Широкий В.Р., Доленко С.А. Прогнозирование Dst-индекса, основанное на методах машинного обучения // Космич. исслед. Т. 56. № 6. С. 353‒364. 2018.
- – Зорич В.А. Многомерная геометрия, функции очень многих переменных и вероятность // Теория вероятностей и ее применения. Т. 59. Вып. 3. С. 436‒451. 2014.
- − Лазутин Л.Л. Мировые и полярные магнитные бури. М., МГУ. 214 с. 2012.
- – Мягкова И.Н., Шугай Ю.С., Веселовский И.С., Яковчук О.С. Сравнительный анализ влияния рекуррентных высокоскоростных потоков солнечного ветра на радиационное состояние околоземного космического пространства в апреле-июле 2010 года // Астрон. вестн. Т. 47. № 2. С. 141–155. 2013.
- − Мягкова И.Н., Широкий В.Р., Владимиров Р.Д., Баринов О.Г., Доленко С.А. Прогнозирование значений геомагнитного индекса Dst при помощи адаптивных методов // Метеорология и гидрология. № 3. С. 38‒46. 2021.
- − Широкий В.Р. Сравнение нейросетевых моделей прогнозирования геомагнитного Dst-индекса на различных наборах данных и сравнение методов оценки качества работы моделей // XVII Всероссийская научно-техническая конференция “Нейроинформатика-2015” с международным участием. Сборник научных трудов. Ч. 2. М., НИЯУ МИФИ. С. 51‒60. 2015.
- − Akasofu S.-I., S. Chapman S. Solar-Terrestrial Physics. Clarendon Press, Oxford. 889 p. 1972.
- − Amata E., Pallocchia G., Consolini G. et al. Comparison between three algorithms for Dst predictions over the 2003–2005 period //J Atmos Sol-Terr Phys. V. 70. P. 496–502. 2008.
- − Barkhatov N.A. et al. Comparison of efficiency of artificial neural networks for forecasting the geomagnetic activity index Dst // Radiophysics and Quantum Electronics. V. 43. № 5. P. 347 ‒355. 2000.
- − Bortnik J., Chu X., Ma Q., Li W., Zhang X., Thorne R.M., Baker D.N. Artificial Neural Networks for Determining Magnetospheric Conditions // Machine Learning Techniques for Space Weather. P. 279–300. 2018.
- − Breiman L., Friedman J.H., Olshen R., Stone C. // Classification and Regression Trees. Wadsworth, Belmont, CA. 1983.
- − Breiman L. Random Forests // Machine Learning. V. 45. P. 5‒32. 2001
- − Burton R.K., McPherron R.L., Russell C.T. An empirical relationship between interplanetary conditions and Dst // J. Geophys. Res. V. 80. P. 4204–4214. 1975.
- − Dolenko S.A., Orlov Yu.V., Persiantsev I.G., Shugai Ju.S. Neural network algorithm for events forecasting and its application to space physics data // Lecture Notes in Computer Science. V. 3697. P. 527‒532. 2005.
- − Friedman J.H. Stochastic Gradient Boosting // Computational Statistics and Data Analysis. V. 38. № 4. P. 367‒ 378. 2002.
- − Haykin S. Neural Networks: A Comprehensive Foundation, 2nd ed. (Prentice Hall, 1998).
- − Kataoka R., Miyoshi Y. Average profiles of the solar wind and outer radiation belt during the extreme flux enhancement of relativistic electrons at geosynchronous orbit // Ann. Geophys. V. 26. P. 1335‒1339. 2008.
- − Lazzús J.A., Vega P., Rojas P., Salfate I. Forecasting the Dst index using a swarm-optimizedneural network // Space Weather. V. 15. P. 1068‒1089. 2017. https://doi.org/10.1002/2017SW001608
- − Lindsay G.M., Russell C.T., Luhmann J.G. Predictability of Dst index based upon solar wind conditions monitored inside 1 AU // J. Geophys. Res. V. 104. № A5. P. 10 335‒ 10 344. 1999.
- − Myagkova I., Shiroky V., Dolenko S. Prediction of geomagnetic indexes with the help of artificial neural networks // E3S Web of Conferences, 20: art. 02011, 2017. https://doi.org/10.1051/e3sconf/20172002011
- − O’Brien T.P., McPherron R.L. Forecasting the ring current index Dst in real time // J. Atmosph. and Sol.-Terrestr. Phys. V. 62. P. 1295‒1299. 2000.
- − Pallochia G. et al. Geomagnetic Dst -index forecast based on IMF data only // Ann. Geophys. V. 24. P. 989‒999. 2006.
- − Patra S., Spencer E., Horton W., Sojka J. Study of Dst/ring current recovery times using the WINDMI model //J. Geophys. Res. V.116. A02212. 2011. https://doi.org/10.1029/2010JA015824
- − Podladchikova T.V., Petrukovich A.A. Extended geomagnetic storm forecast ahead of available solar wind measurements // Space Weather: The International J. Research and Applications. V. 10. CiteID S07001. 2012.
- − Pulkkinen T. Space Weather: Terrestrial Perspective // Living Rev. Solar Phys. 4. 1. URL (cited on 18 September 2007): http://www.livingreviews.org/lrsp-2007-1. 2007.
- − Revallo M., Valach V., Hejda P., Bochníčeket J. Modeling of CME and CIR driven geomagnetic storms by means of artificial neural networks // J. Atm. and Sol. Terr. Phys. V. 110. № 9. 2014.
- − Schrijver, Carolus J. et al. Understanding space weather to shield society: A global road map 772 for 2015–2025 commissioned by COSPAR and ILWS // Adv. in Space Res. V. 55. P. 2745‒2807. 2015.
- − Sugiura M. Hourly values of equatorial Dst for the IGY // Ann. Int. Geophys. Pergamon Press, Oxford. V. 35. P. 9‒45. 1964.
- − Wu J.-G., Lundstedt H. Geomagnetic storm predictions from solar wind data with the use of dynamic neural networks // J. Geophys. Res. V. 102. № A7. P. 14 255‒14 268. 1997.
Supplementary files
