Prediction Of Network Flow Based On Wavelet Analysis And Arima Model

With the rapid development of computer networks, the current network size is extremely large and complex, Web-based applications is growing rapidly. it means that network services more complex, Web services more prone to problems, the performance of the network more vulnerable. In order to provide quality services, network maintenance and management is very important. To accurately predict network traffic on a computer network design and management, conflict control and dynamic bandwidth allocation plays an important role. However, successful traffic prediction can not do without the support of accurate, high-quality traffic model for the design of high-performance network protocols and efficient network topology, for the design of cost-effective network equipment and servers, for accurate network performance analysis and Forecasting, For congestion management and flow balance and improve service quality and so on has very important significance.This paper describes the theoretical basis of wavelet, including wavelet and wavelet transform, and wavelet transform category. Detailed multi-scale analysis and mallat algorithms. Then algorithm analysis of network traffic model, a brief introduction of the Poisson model, Markov model, AR, MA, ARMA model, focused on analyzing ARIMA Model Algorithm.This paper, we combined wavelet transform and ARIMA time series model, established a network traffic prediction model. on the flow time series, using wavelet decomposition, gets detail coefficients and approximation coefficients. on the detail coefficients, applying stationary ARMA model, For the approximation coefficients, the general algorithm treats it as a smooth sequence, but in most cases it is still a non-stationary signal. so I made an improvement here, using ARIMA (p, d, q) model to make it smoothing, then establish ARMA model. Experimental results show that the predictive using this method than the latter.