Abstract: The prediction of stabled time series often use ARMA model, but to the most of the unstabled time series often use the analysis of certainty and randomness time series. It is a waste of the random information with the analysis of certainty, while after the difference of time series, it comes to ARMA model with the analysis of randomness. The paper gives a way to full use of the residual data with SVR model after using ARMA model. Comparing to the single model used to predicting, the method of the mixture of ARMA and SVR model show to be more practical to improve the prediction precision.

Key words: time series, ARMA, SVR, mixture prediction