Abstract: A prediction model was proposed for self-similar network traffic based on EMD (empirical mode decomposition) and RVM(relevant vectors machine).Firstly,the network traffic in slipping window is decomposed into multiple IMF(intrinsic mode function) using EMD,and then RVM is applied to fit high frequent components while ARMA is used to structure prediction model for low frequent components,lastly,all the components’ forecasting result is composed.The experiment indicates that the proposed model can accuratly predict traffic time series’ amplitude and trend,and compared to other method,achieves higher prediction accuracy than that of other similar prediction methods.

Key words: Network traffic,Prediction model,EMD,RVM