摘要
采用小波神经网络对网络流量数据的时间序列进行建模与预测。针对BP神经网络预测准确率不太理想的情况,将小波理论引入BP神经网络,引用小波理论中多分辨分析技术对基于BP神经网络的模型进行改进,建立了基于小波神经网络的IP网络流量预测模型。该模型利用小波多分辨分析分解信号,再用已分解的信号序列来训练BP神经网络。实验结果表明,小波神经网络比BP神经网络对网络流量的预测结果精度更高、性能更好,利用小波神经网络预测网络流量是一种可行、有效的方法。
The time series of network traffic data is modeled and forecasted based on wavelet neural network.According to the situation that the network prediction based on BP neural network is not well exact,the wavelet theory was introduced into BP neural network.On the basis of wavelet theory in reference multi-solution analysis techniques to improve the model based on BP neural network,the IP network traffic forecasting model based on wavelet neural network was built.We used wavelet multi-solution analysis techniques to decompose the traffic signal and then employed the decomposed component sequences to train the BP neural network.The results of the experiment prove that the wavelet neural network is superior to the BP neural netw ork method in prediction performance.And wavelet neural network is the suitable and effective method for forecasting Internet traffic.
出处
《计算机科学》
CSCD
北大核心
2011年第B10期296-298,330,共4页
Computer Science
关键词
IP网络
流量预测
服务质量
BP网络
小波神经网络
多分辨分析
时间序列
IP network
Prediction of traffic
Quality of service
BP network
Wavelet neural network
Multi-solution analysis
Time series
