Research on the Network Traffic Time Series Modeling and Forecasting Based on Wavelet Decomposition

Abstract A network traffic time series analysis and forecasting method based on wavelet decomposition is proposed in the paper. Firstly, non-stationary network traffic time series are decomposed into multiple stationary components, then various stationary components are modeled using the autoregressive moving average method and at last models of all components are composed to get the prediction model for original non-stationary network traffic time series. In the simulation experiment, we used time series data from the network traffic library to establish a prediction model and carried out independent test and inspection. Simulation test results show that the forecasting method proposed in this paper has improved forecasting accuracy. It is an effective and robust network traffic forecasting method.