The Research Of Traffic Analysis And Forecasting Method For Wireless Network

In current days, the wireless network as new network architecture, more and more areas applied it, such as campus wireless network, military field, medical field and other areas. Following the wireless network development, the wireless network management and network security have coming. In particular, network traffic analysis and forecasting for network traffic is one of the important solutions for network management and security. This paper analyzed the character of the wireless network and studied the different forecasting methods. Base on research, this paper proposed a traffic forecasting method which fit the character of the wireless traffic. In the following, there are several studies presented in this paper.1) The analysis of the wireless traffic. In this part, the analysis was mainly to capture the character of the network traffic. On the one hand, did the data analysis in order to study characterization of the traffic experiment data-set. In the other hand, we did the statistic analysis of the wireless network traffic using the probability statistics knowledge, attached the distribution characterize, which will be founded for building the new forecasting model in the next part.2) The research of the wireless traffic forecasting model. In this research, a new network traffic forecasting model was proposed, named GCSVR. As a new mathematical method, we employed the Chaos theory to analysis and phase space reconstruction the traffic time series; and smooth the time series using the Grey theory, the intent is to make the stochastic of the raw traffic time series weaken, get a smooth time series which have a fully characterize and strong regularity. At the last of the model, forecast the wireless traffic using the Support Vector Regression (SVR). In the experiment part, we did some experiments using the proposed model which was compared with the original SVR. The experiments have proved that the GCSVR model has a high capability, it can handle the unstable time series which have a highly burst and an awful regularity particularly; it can smooth the raw time series into a smooth series; it also have a high forecasting accuracy and have a far forecasting step.