Predicting the corrections for the polish timescale UTC(PL) using GMDH and GRNN neural networks

Abstract

The article presents the results of research on the prediction of the polish timescale UTC(PL) based on GMDH and GRNN neural networks, which were compared with the results obtained in the GUM using the analytical linear regression method. The lowest values of the prediction error was obtained for GMDH neural network for time series analysis methods and data prepared on the basis of time series ts1. These results were significantly better than the prediction error values obtained in GUM using analytical linear regression method. In the case of GRNN neural network prediction errors obtained using the regression method and data prepared on the basis of time series ts2 are very close to the values of prediction error obtained in the GUM. However, for data prepared on the basis of time series ts1 reached a very high value.