This paper proposed a methodology of forecasting for mobile communication traffic with maximal overlap discrete wavelet transform (MODWT) according to priori knowledge.Fourier spectrum was chosen as the priori knowledge to avoid the blindness of wavelet decomposition.Then

MODWT which is easy to extract components with obvious physical meaning was employed to decompose the communication traffic data.Moreover

prior knowledge of fourier spectrum was taken as reference to synthesize relevant sublayers

leading to the trend and seasonal components.Further

seasonal autoregressive integrated moving average (ARIMA) model was applied to model and predict the previous trend and seasonal components

respectively.The results tested with real communication traffic data indicate:the methodology proposed in this paper can realize multistep prediction and the forecasting accuracy is superior to that of seasonal ARIMA models.