Forecast daily indices of solar activity, F10.7, using support vector regression method

Abstract

The 10.7 cm solar radio flux (F10.7), the value of the solar radio emission flux density at a wavelength of 10.7 cm, is a useful index of solar activity as a proxy for solar extreme ultraviolet radiation. It is meaningful and important to predict F10.7 values accurately for both long-term (months-years) and short-term (days) forecasting, which are often used as inputs in space weather models. This study applies a novel neural network technique, support vector regression (SVR), to forecasting daily values of F10.7. The aim of this study is to examine the feasibility of SVR in short-term F10.7 forecasting. The approach, based on SVR, reduces the dimension of feature space in the training process by using a kernel-based learning algorithm. Thus, the complexity of the calculation becomes lower and a small amount of training data will be sufficient. The time series of F10.7 from 2002 to 2006 are employed as the data sets. The performance of the approach is estimated by calculating the norm mean square error and mean absolute percentage error. It is shown that our approach can perform well by using fewer training data points than the traditional neural network.

A Neural Network‐Based Ionospheric Model Over Africa From Constellation Observing System for Meteorology, Ionosphere, and Climate and Ground Global Positioning System Observations

“…We emphasize that our reason to use the F10.7 is based on the observation that it gave the least RMSE and MAE for the data set used in this work. Some authors (e.g., Huang et al, 2009;Antony & Shanthi, 2015; and in the NeQuick model) have also shown that the F10.7 is a good proxy for solar activity studies in the upper atmosphere.…”

Effects of solar wind ultralow‐frequency fluctuations on plasma sheet electron temperature: Regression analysis with support vector machine

“…(4) It avoids traps by local minima. In space physics, SVM has been used in forecasting, including solar F 10.7 [ Huang et al , ], solar wind velocity [ Liu et al , ], substorms [ Gavrishchaka and Ganguli , ], Kp [ Ji et al , ], and storm time ionosphere [ Sun et al , ]. SVRM has been applied to model the magnetopause shape [ Wang et al , ].…”

Forecasting F10.7 with solar magnetic flux transport modeling

“…To a large extent these fluxes control conditions in the ionosphere that play a major role in communications and navigation signals, e.g., of the ionospheric total electron content [e.g., Richards et al , 1994; Maruyama , 2010]. Current UV and EUV forecast models commonly rely on auto‐regressive and time series analysis of past solar measurements of, e.g., the sunspot number (SSN) and solar 10.7 cm (2.8 GHz) radio flux [e.g., Chatterjee , 2001; Lean et al , 2009; Huang et al , 2009], commonly abbreviated as F 10.7 . Parameters such as SSN and F 10.7 are, in general, proxies for solar magnetic activity.…”