在无抽取Haar小波变换的基础上,结合自适应AR模型和滑动窗口式多项式拟合方法,建立了一种基于小波变换的递推式高速网络流量在线预测模型。该模型首先用无抽取Haar小波变换把网络流量时间序列分解为细节信号和近似信号,然后对细节信号部分使用自适应AR模型预测,对近似信号部分则使用滑动窗口式多项式拟合方法预测,最后用小波重构获得原始时间序列的预测值。该模型不但提高了流量在线预测的准确性,而且通过模型参数的递推式自动调整,避免了参数的定期估计和更新。
Based on non-decimated Haar wavelet transform, a recursive prediction model of network traffic is proposed. This model decomposes the network traffic time series with non-decimated Haar wavelet transform firstly, then predicts approximate signal with AAR model and predicts detailed signal with sliding window polynomial fitting, re-composes the prediction value of original series finally. This model not only improves the accuracy of network traffic prediction, but also implements the online update of model parameters by recursive estimation.