Ensemble Machine Learning of Random Forest, AdaBoost and XGBoost for Vertical Total Electron Content Forecasting

Space weather describes varying conditions between the Sun and Earth that can degrade Global Navigation Satellite Systems (GNSS) operations. Thus, these effects should be precisely timely corrected for accurate reliable GNSS applications. That modeled with Vertical Total Electron Content (VTEC) in Earth’s ionosphere. This study investigates different learning algorithms to approximate nonlinear space processes forecast VTEC 1 h 24 future low-, mid- high-latitude ionospheric grid points along same longitude. models are developed using of Decision Tree ensemble Random Forest, Adaptive Boosting (AdaBoost), eXtreme Gradient (XGBoost). Furthermore, combined into a single meta-model Voting Regressor. Models were trained, optimized, validated time series cross-validation technique. Moreover, relative importance input variables is estimated. The results show perform well both quiet storm conditions, where multi-tree outperforms Tree. In particular, meta-estimator Regressor provides mostly lowest RMSE highest correlation coefficients as it averages predictions from well-performing models. expanding dataset derivatives, moving averages, daily differences, modifying data, such differencing, enhances features, especially over longer horizon.