Abstract:Aiming at the poor performance of short term prediction of navigation satellite clock error, a variable weight combined ARIMA and ANN method is proposed, which combines the virtues of ARIMA and ANN. To improve the accuracy of prediction, an additional momentum term is used to modify the weight of the neural network and a sequence relative nearness degree is used to modify the weight of models. The clock data of 4 typical GPS satellites are chosen and respectively used in ARIMA, ANN and variable weight combination models to forecast short term clock error. The results show that the accuracy of variable weight combination model is superior to the other models, especially in the field of rubidium clock, the error of which is half of the other models.