Analyzing And Forecasting Of 3G Data Services Based On Service Models

Since WCDMA, CDMA2000, TD-SCDMA 3G networks were put into operation, mobile data services have been more and more widely used. In the meantime, operators are increasingly concerned about how to improve the quality of service in networks with limited wireless resources, so that they could offer better user experience and furtherly enhance themselves both in competitiveness and economic efficiency. Network planning and optimizing aims to improve the quality of service, and approaches and tools of analyzing and forecasting of data services contribute to efficient network planning and optimization.The mobile users demonstrate the same or similar behaviors of using data services when they are in the same or similar geography or time scenes, which makes the traffic of data service change in similar patterns. The users’ behavior of using data services can be charaterized with a series of indicators, such as the uplink and downlink traffic generated by each usage of a data service, the number of usages of a data service in unit time. For instance, the service behaviors of users watching matchs in different stadiums are similar, and so are the ones on Spring Festival and New Year’s Day.Taking into account the similarity of service usage behavior in similar scenes, this paper establishes several types of data service models to describe the service usage behavior of different user groups in different geography or time scenes. The main indicators of these service models include the mean and variance of the uplink/downlink traffic generated by each service usage, the calling/called number of a service usage per user in unit time, the number of online subscribers and service permeability in the period during which the model is established, etc.The similarity trees of geography or time scenes are defined, which provide the guideline for query of service models in the same or similar scenes to forecast different types of data services. There are three kinds of forecasting modes, that is, the mode for holiday and special periods when the number of subscribers can be estimated, the mode for holiday, special periods and unknown traffic-bursting periods when the number of subscribers cannot be estimated and the mode for regular periods. Furtherly, with the analysis and calculation of the indicators such as the mean and variance of the uplink/downlink traffic generated by each service usage and the calling/called number of a service usage per user in unit time, the predicting results in terms of the mean and maximum of the service uplink/downlink traffic in unit time can be achieved. These results provide operators with a decision making basis for the optimization and allocation of wireless network resources.In the mode for regular period forecasting, the Holt-Winters model is used for predicting the number of users and service calls.Finally, a prototype for analyzing and forecasting of 3G data services is developed based on the methods mentioned above. The system is developed by means of C++, running in.NET platform and SQL Server 2005 database. The system is tested and verified by the practical service data, and the results indicate that it achieves the desired goal of analysis and forecast.