Self-similarity Analysis and Application of Network Traffic

Network traffic prediction is not only an academic problem, but also a concern of industry and network performance department. Efficient prediction of network traffic is helpful for protocol design, traffic scheduling, detection of network attacks, etc. In this paper, we propose a network traffic prediction method based on the Echo State Network. In the first place we prove that the network traffic data are self-similar by means of the calculation of Hurst exponent of each traffic time series, which indicates that we can predict network traffic utilizing nonlinear time series models. Then Echo State Network is applied for network traffic forecasting. Furthermore, to avoid the weak-conditioned problem, grid search algorithm is used to optimize the reservoir parameters and coefficients. The dataset we perform experiments on are large-scale network traffic data at different time scale. They come from three provinces and are provided by ZTE Corporation. The result shows that our approach can predict network traffic efficiently, which is also a verification of the self-similarity analysis.