GNSS时间序列多尺度分解与预测软件设计OA

To address the cumbersome processing of global navigation satellite system(GNSS)time series data and the need for comparative analysis among multiple algorithms,a GNSS coordinate time series prediction software called GNSS_pre was designed,based on matrix laboratory(MATLAB).This software integrates decomposition models such as empirical mode decomposition(EMD)and prediction models including backpropagation(BP)neural networks.It constructs modules for time series input,decomposition,prediction,and accuracy evaluation,processing data through a standardized computa-tional workflow while offering data visualization,storage,and module expansion capabilities.Validation experiments using daily time series data from ten GNSS stations demonstrate that the software operates stably and reliably,with strong predic-tion accuracy,robustness,and fitting ability across various models.Specifically,the complete ensemble empirical mode decomposition with adaptive noise-extreme learning machine(CEEMDAN-ELM)model outperforms in four accuracy met-rics:root mean square error,mean absolute error,coefficient of determination,and mean bias error.The software's modu-lar design and reserved expansion slots provide new ideas and valuable references for flexible combinations of models in the field of GNSS time series prediction.