Abstract: Based on the phase space reconstruction improved by wavelet transformation, this paper takes influence of wavelet denoising into account, and introduces the method of chiastic time series predictor based on wavelet neural network traffic flow prediction. Experimental results show that, comparing with RBF neural prediction method, wavelet neural network prediction method approaches better result, and the error margin is smaller.

Key words: network traffic flow, phase space reconstruction, wavelet neural network

摘要： 在Takens提出的相空间重构模型基础上，应用小波变换对其进行改进，充分考虑噪声对重构结果的影响。将小波神经网络混沌时间序列预测方法引入网络流量预测中，介绍小波神经网络的基本构造和学习方法。实验表明，与RBF神经预测方法相比，小波神经网络预测方法的逼近效果更好、误差更小。

关键词: 网络流量, 相空间重构, 小波神经网络

CLC Number: 

• TP393