GNSS (global navigation satellite system) vertical time series have the characteristics of non-stationary, non-linear, and noisy. Based on the in-depth study of the Prophet prediction model, and the good predictive effect of Prophet prediction model on trend signals and periodic signals, a “noise reduction-decomposition-prediction” combined prediction method of GNSS vertical time series that introduces EMD (empirical mode decomposition) was proposed. EMD denoising was performed on the original time series, the denoised series were decomposed and predicted, and the predicted signal of each component was reconstructed into the final predicted series. The measured vertical data was used for research, and results show that the average signal-to-noise ratio of the signal after noise reduction is 10.30 dB, and the average energy percentage is 88.75%; using the short-term prediction method, the root-mean-square errors of GNSS vertical time series prediction results are increased by 26.41% and 14.88% on average, respectively; the average percentage errors are increased by 18.92% and 7.91% on average, respectively, and the effectiveness and practicability of the combined forecasting method are verified.

The study of GNSS (global navigation satellite system) vertical time series is helpful for monitoring and analyzing the movement of crustal plates, and can provide an important basis for judging the movement trend. A MEMD-XGBoost model was constructed based on empirical mode decomposition and extreme gradient boosting algorithm for GNSS vertical time series prediction and analysis. In order to verify the prediction performance of the model, the vertical time series data of 8 GNSS stations were selected for prediction experiments. The feature construction results show that multiple empirical mode decomposition can accurately extract the original time series information and provide effective features. The modeling results show that the MEMD-XGBoost model can effectively improve the data quality. The prediction results show that the prediction results of the MEMD-XGBoost model have high precision and accuracy, and the degree of error dispersion is small, the model has strong stability and robustness, and can better predict the movement trend and seasonal changes in the U direction of the GNSS station. Therefore, the model can be applied to GNSS vertical time series modeling and prediction research.