Abstract: This article proposes a new method to model and predict network traffic based on fractal filter and linear neural network.A smoothing series can be generated after increasing the long-range dependence in the time series by fractal filter，according to the theory of phase space restructure，the linear neural network model is constructed，experiments of real network traffic illustrate that the prediction results using the method are better than the traditional models.

摘要： 提出了一种结合分形滤波与线性神经网络进行网络流量预测的新方法。通过分形滤波增强网络流量中的长相关结构，使序列更加平滑，根据相空间重构理论利用线性神经网络进行预测操作，并用实际网络流量验证该模型的有效性。