基于小波包的RBF神经网络网络流量混沌预测

摘要

In order to improve the accuracy of the network traffic prediction, a method of RBF (radial basis function, RBF) neural network chaotic prediction of network traffic based on wavelet packet is proposed. Besides, when calculating the optimal embedding dimension and time delay in WPCRBF, an improved method of time window is applied with consideration of the effect of nose and the periodic characteristic of network traffic The proposed method of WPCRBF is tested on the prediction of real network traffic. And then compared its predictive performances with CRBF's and wavelet-based RBF neural network of chaotic prediction method WCRBF's. Experimental results show that the proposed method perform the best predictive accuracy, and WPCRBF can server as a promising alternative for the complicated network traffic prediction.%为了提高网络流量预测准确率,提出了基于小波包的RBF神经网络网络流量混沌预测法(WPCRBF).充分考虑到真实网络流量的周期性和噪声的影响,提出了一种改进的时间窗口法来计算最佳嵌入维和时间延迟,并用于上述预测方法中.以真实网络流量数据为实验数据,分别用CRBF、基于小波的RBF神经网络混沌预测法(WCRBF)与提出的WPCRBF进行预测,实验结果表明,该方法能够较准确地对网络流量进行预测.