Research Of Data Traffic Prediction For Wireless Communications

With the rapid development of wireless communication technology, the improvement of data transmission speed, and especially the popularity of the3G technology, the data traffic is increasing dramatically. However, the deficient traffic prediction causes the lack of effective guidance on network construction and adjustment, which affect the experience of uses and enterprise revenue. Therefore, the accurate data traffic prediction has become urgent issue in the wireless communication field. This thesis focuses on the prediction process of traffic data based on the characteristics of the data traffic, so that the relevant models can be applied to the data prediction.The thesis briefly introduces the basic knowledge of time series and prediction. Then, the process of prediction is minutely showed, including the model choice, order and parameters estimation, model test and finally error analysis. The key of the thesis is seasonal time-series model and the ARCH model dealt with conditional variance. Also, two other models are applied to fit the data, in order to compare the fitting results with each other.The thesis first analyzes the statistics characteristics based on the methods of voice traffic processing, such as, spectrum analysis, seasonal property, stability and pure random sequence, and so on. After that, according to the data related properties, a series of meticulous preprocessing is carried out. Then try to establish classic seasonal time series model and choose the most suitable parameters to make the prediction error as little as possible. Later, the residual series of the time series model are found to have heteroscedastic property. Then ARCH model (Autoregressive Conditional Heteroskedasticity Model) is adopted to fit the data traffic, and the model is called Season-ARMA-ARCH. By comparing to the result of traditional time-series model, Season-ARMA-ARCH model improved the prediction precision.All in all, the core of this thesis is to realize the application of data traffic prediction. And the key is the design of prediction system. By studying the known data, the scattered prediction theories are integrated to complete the practical data prediction application.