Forecasting Sunspot Time Series Using Deep Learning Methods
To predict Solar Cycle 25, we used the values of sunspot number (SSN), which have been measured regularly from 1749 to the present. In this study, we converted the SSN dataset, which consists of SSNs between 1749 – 2018, into a time series, and made the ten-year forecast with the help of deep-learning (DL) algorithms. Our results show that algorithms such as long-short-term memory (LSTM) and neural network autoregression (NNAR), which are DL algorithms, perform better than many algorithms such as ARIMA, Naive, Seasonal Naive, Mean and Drift, which are expressed as classical algorithms in a large time-series estimation process. Using the R programming language, it was also predicted that the maximum amplitude of Solar Cycle (SC) 25 will be reached between 2022 and 2023.
KeywordsSunspot number Statistics Solar cycle
Disclosure of Potential Conflicts of Interest
The authors declare that they have no conflicts of interest.
- Adhikari, R., Agrawal, R.K.: 2013, arXiv.
- Dancho, M.: 2018, blogs.rstudio.com/tensorflow/posts/2018-06-25-sunspots-lstm/.
- Lewis, N.D.: 2016, Deep Time Series Forecasting with Python: An Intuitive Introduction to Deep Learning for Applied Time Series Modeling, Create Space Independent Publishing Platform, New York. Google Scholar
- Penn, M.J., Livingston, W.: 2011, In: Choudhary, D.P., Strassmeier, K.G. (eds.) The Physics of Sun and Star Spots, IAU Symp. 273, 126. Google Scholar
- Sena, D., Nagwani, N.K.: 2016, J. Eng. Appl. Sci. 11, 13123. Google Scholar
- Siami-Namini, S., Namin, A.S.: 2018, arXiv.