An analysis of the combined wavelet-GM (1,1) model for network traffic forecasting

Abstract

Based on Grey theory, a combined wavelet-GM (1,1) forecasting model is proposed. The random properties of some non-stationary time series can be reduced by wavelet decomposition into many series according to different scales. Decomposed time series are predicted with GM (1,1) model to obtain forecasted results of the original time series. Experiments on network traffic show that the combined model is more effective than the common grey model. The application of the proposed combined model to network traffic forecasting may offer scientific rationale to Internet routing decision, LAN virus detection and illegal invasion prevention.