Determination of Optimal Maximum Usable Frequency Prediction Periods for Ionospheric Radio Channels Using the XGBoost Machine Learning Algorithm

An algorithm has been developed for applying the XGBoost machine learning method to solve the problem of predicting the maximum usable frequencies of shortwave communication channels with and without hyperparameter tuning. The optimal (minimum and maximum possible) periods for predicting the maximum usable frequencies are determined based on training samples compiled from radio sounding data of short-wave communication channels obtained using Ionosonde VSUT equipment. A database of machine learning models XGBoost was obtained for various combinations of hypermeters and forecast periods. The effectiveness of the development was verified for a mid-latitude radio path with a length of 2600 km using the MUF (Maximum usable frequency) dataset for thirty days.