Efficiency of the least squares support vector regression in local modeling of the ionosphere total electron content and comparison with other models

Abstract

In this paper, the aim is to use the least squares support vector regression (LS-SVR) for spatio-temporal modeling of the ionospheric total electron content (TEC). In order to do this, the observations of 15 GPS stations in the north-west of Iran have been used in the period from 193 to 228 at 2012. Comparing the results of the new model with support vector regression (SVR), artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), Kriging model, GIM and international reference ionosphere 2016 (IRI2016) as well as TEC obtained from GPS. The analyzes performed show that the averaged RMSE of ANN, ANFIS, SVR, LS-SVR, Kriging, GIM and IRI2016 models in two interior control stations are 3.91, 2.73, 1.27, 1.04, 2.70, 3.02 and 6.93 TECU, respectively. Also, the averaged relative error of the models in two interior control stations was calculated as 15.98%, 9.39%, 7.85%, 6.09%, 11.60%, 12.54% and 26.56%, respectively. Analysis of the PPP method shows an improvement of 50 mm in the coordinate components using the LS-SVR model. The results of this paper show that the LS-SVR model can be considered as an alternative to global and empirical models of the ionosphere in the study area.

Keywords

• ionosphere
• tec
• gps
• northwest of iran
• machine learning
• ls-svr