Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{181516,
        author = {Adrika Kakoty and Lipika Soni and Barsha Dutta and S. Aruna},
        title = {Prediction of Total Electron Content by using Machine Learning Approach over Low Latitude Regions during Low Solar Activity Period},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {1},
        pages = {4949-4961},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=181516},
        abstract = {Total Electron Content (TEC) present in the ionosphere plays a crucial role in the transmission of signals through the ionosphere. So, the study of ionospheric TEC, its variations (diurnal, monthly, seasonal, yearly, etc.), and the various effects like solar activity, geomagnetic activity, etc. on TEC becomes an important topic for navigation. The machine learning methods are very useful to solve non-linear prediction challenges, especially for short-term GPS TEC forecasting due to the composite spatio-temporal variations. This study analyzes the performances of seven different machine learning models viz, RandomForest, XGBoost, GradientBoosting, LightGBM, LSTM, GBDT, GRU, and compares them to predict the ionospheric TEC over two low-latitude stations Pathum Wan (13.74° N, 100.54° E) and Bangalore (12.97° N, 77.59° E) separated by 23° longitude in the year 2009, low solar activity period (65.8 < F10.7 < 86.9). The model outputs have been validated by the real time data over the two locations. Solar quiet period is selected for the better analysis of day-to-day changes of the ionosphere which is important to improve the accuracy of physics-based models of the ionosphere. The monthly average TEC values predicted by all the models show strong agreement with the observed TEC values at both stations, with correlation coefficients 0.88 < R2 < 0.93 over Pathum Wan and 0.85 < R2 < 0.92 over Bangalore. The Root Mean Square Errors (RMSE) in the correlation plots are found between 2.13 to 2.87 over Pathum Wan and 2.29 to 3.25 over Bangalore. The values of Mean Absolute Errors (MAE) over these two locations are in between 1.48 to 2 and 1.51 to 2.27 respectively.  Among all seven models LightGBM (R2 = 0.926 over Pathum Wan and 0.918 over Bangalore) and XGBoost (R2 = 0.923 over Pathum Wan and 0.914 over Bangalore) show the best correlations with lowest RMSE (2.1 over Pathum Wan and 2.3 over Bangalore) and MAE (1.5 over both Pathum Wan and over Bangalore).},
        keywords = {TEC, VTEC, Ionosphere, Machine Learning, RMSE, MAE, Correlation},
        month = {June},
        }