Abstract: The time series of network traffic data is modeled and forecasted based on wavclet 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 wavclet multi-solution analysis techniques to decompose the traffic signal and then employed the decomposed component sectuences to train the I3P 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 wavclet neural network is the suitable and effective method for forecasting Internet traffic.

Key words: IP network, Prediction of traffic, Quality of service, BP network, Wavelet neural network, Multi-solution analysis, Time series