The Research Of Dependence In The Internet And Network Traffic Prediction

The popularity of Internet and video-on-demand lead to sharp increase in network traffic, at this time the traditional characteristics of network traffic are no longer applicable to network traffic prediction. After measuring network traffic of local area network and Internet, researchers find that Internet traffic has self-similarity of statistically significant, because the feature of self-similarity can describe network traffic more accurately, therefore it is more suitable of network traffic prediction than traditional characteristics of network traffic.Currently, the growing scale of network is a challenge for reliable transmission of data and network resource allocation, and there comes more and more network congestion and failures. Accurate prediction of network traffic has important reference value for removing of network congestion and management of complex network, so it becomes increasingly urgent to realize high quality network traffic prediction.In order to find a new network traffic prediction model as well as to improve the accuracy of network traffic prediction, this paper firstly introduces the research status of self-similarity and network traffic prediction, and then details the self- similarity and traditional network traffic prediction model, in the end improves the network traffic prediction model and proposes a combination network traffic prediction model. Experimental results show that this model is feasible.A BP neural networks learning algorithm based on the improved differential evolution is proposed in this paper and the algorithm is a stochastic global optimization technique. Using the ability of global optimization, the algorithm is the one best in training the weight values and threshold values of BP networks. The experiments results show that the proposed algorithm speeds up the convergence rate of BP neural network and the accuracy of the network traffic prediction is improved.