基于小波分解的网络流量时间序列建模与预测

Modeling and forecasting for network traffic based on wavelet decomposition

提出一种基于小波分解的网络流量时间序列的分析和预测方法。将非平稳的网络流量时间序列通过小波分解成为多个平稳分量,采用自回归滑动平均方法分别对各平稳分量进行建模,将所有分量的模型进行组合,得到原始非平稳网络流量时间序列的预测模型。在仿真实验中,利用网络流量文库的时间序列数据建立了预测模型,并对其进行独立测试检验。仿真结果表明,本预测方法提高了网络流量时间序列的预测准确率,是一种有效、稳健的网络流量预测方法。

This paper proposed a network traffic forecasting methods based on wavelet decomposition and time series analysis method. Firstly, the method decomposed the network traffic time series in multiple stationary components by wavelet decomposition, then used the autoregressive moving average method to model the each stationary component separately. Finally combined all the components of the model to get the forecasting model of the original non-stationary network traffic time series. It carried out the simulation experiment on time series data of the network library. The simulation results show that, the proposed method improves the network traffic time series forecasting accuracy rate, and it is an efficient, robust network traffic forecasting method.