Abstract: The earth's rotation parameters(ERP)is an essential parameter for the conversion between the geocentric celestial reference system(GCRS)and the international terrestrial reference system(ITRS).It is the important products of the International GNSS Service Organization(IGS)and the International GNSS Monitoring and Evaluation System Analysis Center(iGMAS).In this paper,the least squares earth rotation parameter prediction algorithm will cause the data saturation and the old and new data in the data processing and forecasting are equally treated and so on,the forgetting factor into the least squares prediction algorithm,and thus improve the accuracy of ERP forecast.The forgetting factor recursive least squares algorithm can prevent data saturation,reduce the influence of old data,strengthen the function of new data,reduce the probability of inversion matrix of rank loss matrix when solving the fitting parameters,and improve the prediction accuracy.In this paper,the least squares expression of forgetting factor is deduced in detail,and the best forgetting factor is explored.The experimental results of this method and the results of LS-AR model are compared with those of LS-AR model.The least squares model prediction of the forgetting factor can achieve the same accuracy as the LS-AR model.

Key words: earth rotation parameters, forecasting, forgetting factor, the recursive least squares