The Study of Machine Learning Features to Analyze the Accuracy of Prediction of the Coherence Band of GLONASS Reference Stations

To improve the prediction of the values of coherence bands in the transionospheric communication channel, a technique was developed that uses data mining methods and the values of reference stations. For this, machine learning (ML) models based on the AdaBoost, Gradient boosting and XGBoost algorithms were used. Based on the test results, it was revealed that the Gradient Boosting model turned out to be the most accurate with an average error of 13.4 MHz and an accuracy of about 90% on a 60-day dataset within the confidence interval.