Traffic Prediction Model Based On Wavelet And Markov

With the rapid development of network technology and the increasing of network-scale, and as well as the application of network become more and more diverse, the traditional network management had not meet the higher requirements for QoS of the users. The analyzing characteristic of network traffic is not only the base to prompt network performance, to optimize the network topology and to balance the network load, but also is the important means of finding web service abnormity and network failure. Therefore, the modeling and prediction of network traffic is of great significance to design large-scale network plan, manage network resource, adjust user behavior and detect network anomalies and failures。Currently, in-depth study of network traffic characteristics found that self-similarity is considered the most important statistical characteristics of network traffic, such as long-range dependence, Continuous burst and Multifractal, etc. This feature not only exist in the Internet networks, but also exist in wireless networks, AdHoc networks and satellite networks. The behavior of network traffic characteristics of this complex is usually expressed in most of the time scale and statistical order on the sudden, so this makes the traditional Poisson process, Markov process and other stochastic models were not suited to the description of the current and forecast network traffic. Therefore, the self-similar network traffic modeling and prediction play a very important role in network capacity planning and ensuring the performance of heterogeneous network.After analyzing the wavelet theory and the Markov chain, a network traffic prediction model based on wavelet transform and Markov chain is proposed. The model employs wavelet transform which decomposes the traffic into high frequency coefficients and low frequency coefficients, and then these different frequency coefficients are reconstructed by single branch to the high frequency traffic parts and the low frequency traffic parts which are sent individually into different Markov chain for prediction. The synthesizer of the Markov chain outputs are the predicted results of the original network traffic.Adopted Qingdao Agricultural University campus network traffic data as a the testing object, single step and multi-step traffic prediction experiment are complete by this prediction model, and compare with several existing models of network traffic. The experimental results prove the efficiency and thesuperiority of the proposed prediction model.Finally, the main contributions in this dissertation are summarized and some suggestions and directions for the future work in this field are given.