Prediction of Trends of Ionospheric and Transionospheric Communication Channel Characteristics by XGBoost Method

A methodology for predicting the characteristics of short-wave (SW) and satellite communication systems using the same hyperparameter selection algorithm of the XGBoost machine learning model with identical starting parameters of the hyperparameter search space is investigated. The validation of the methodology was performed using the maximum usable frequency data for SW communications and the coherence bandwidth for satellite communications. The accuracy of the predictions was evaluated using the relative metrics of coefficient of determination and MAPE.