Machine Learning for GNSS Time Series Analysis in the Time Domain

The paper presents the results of developing a method for analyzing time series of GNSS measurements based on a machine learning approach. The constructed algorithm was tested on GNSS data from the vicinity of sources of large earthquakes occurred in regions with different tectonic structures: the Japanese islands, Southern California, and the Peruvian-Chilean coast. It is shown that the proposed approach allows one to build an adequate, versatile, interpretable, statistically significant time series model using exclusively statistical data analysis methods, which will further allow one to create automated processing systems operating in a near-real-time mode.