Analysis and Prediction of TEC Variations Using Different ML Algorithms

To estimate the local ionosphere's total electron concentration (TEC) with high precision in order to facilitate the use of satellite location, measurement controlling and navigation, we try to explore an empirical machine learning models for predicting VTEC, or vertical total electron content, on a single station receiver located in Bangalore (IISC) 13.02° N &: 77.57°, which is significant for monitoring space weather and ionospheric irregularities. In this analysis GNSS derived Vertical TEC (VTEC) predicted over IISC using 04 different machine learning models, including Bagged Trees, Random Forest Regression, XG Boost, and K-Nearest Neighbors. These models were evaluated using RMSE and R2 metrics. This study effectively predicted VTEC values and the results obtain is XG Boost and Bagged Trees are performed well with an Root mean square error (RMSE) value of 3.034 and 3.035 respectively and R2 value is of 0.985 and 0.984 respectively. Comparing to other models like KNN and Random forest were not much effective as XG Boost and bagged trees. This models when compare to other deep learning approaches like Recurrent Neural Networks (RNNs) which are complexity in nature, requires more data and also consumes more time along with high GPU is required for training. Accurate TEC prediction is therefore essential for reliable GNSS applications and seism ionospheric studies.