ПРОГНОЗИРОВАНИЕ ПОЛНОГО ЭЛЕКТРОННОГО СОДЕРЖАНИЯ ИОНОСФЕРЫ НА ОСНОВЕ АЛГОРИТМОВ МАШИННОГО ОБУЧЕНИЯ

Abstract

Introduction. The paper presents and describes the technology of applying the machine learning algorithm in predicting the vertical total electron content of the ionosphere. The ionospheric error is one of the most significant sources of pseudorange measurement errors from GNSS signals. Increasing every year requirements for the accuracy of positioning and navigation by GNSS signals leads to the need to develop new methods to reduce the impact of various measurement errors, including the ionospheric error. At present, ionospheric models of various types are used for ionospheric correction of measurements. The currently widely used ionospheric models do not allow a significant increase in the accuracy of positioning based on GNSS signals. At the moment, the creation of a new effective method for modeling and forecasting the ionosphere that meets modern requirements for positioning accuracy is an important and urgent task. The purpose of this work is to create a methodology for modeling and predicting the total electron content of the ionosphere using machine learning algorithms. Machine learning is currently a fairly common and popular method for solving problems of classification, recognition and prediction. The method has been used for many years in medicine, robotics, industry, finance and many other branches of modern science and economics. To achieve this goal, it is necessary to solve a number of tasks. First of all, you need to select and collect data for training the model, then you need to select a machine learning method and hyperparameters for the selected method. Next, it is necessary to perform TEC prediction based on the trained model and evaluate the accuracy of the results obtained. comparison of the obtained results with the accuracy of other existing models It is shown that machine learning does a good job of predicting full electronic content. The resulting trained model makes it possible to obtain a forecast with an accuracy comparable to the accuracy of the Klobuchar, NeQuick models, and in some cases much more accurate.