Mobile Communication Data Service Traffic Analysis And Forecasting

Recently, the rapid development of Mobile Communications Technology and Network in our country has turned the wireless communication into the new focus.The wireless network planning and optimization as a significant process of the Wireless network engineering and construction get more and more attention.3 G has offered larger capacity and more flexible high-rate and multi-rate data transmission ability.So, this new situation will set a still higher demand on the planning of the wireless network.The telephone traffic is an important part of the wireless network planning,which target is to forecast the future traffic scientificly and accurately and to provide basis for the wireless planning.Meanwhile, because of the imbalance of economic development and the differences of district circumstances and other reasons, the telephone traffic has the feature of unbalance development and greatly liquidity. While in the same geography scene or a certain period of time, there is some similarity of the traffic bursts. Therefore, based on the historical traffic characteristics, the similar scene trees are built to match the future traffic burst scene with the historical ones and to raise forecasting accuracy.A method of forcasting the telephone traffic of the mobile communication network based on the similar scene trees is presented, which regards the historical traffic data as a kind of time series data and models the data using the time series decomposition method to forecast the future traffic.And the similar scene trees are built as templates for the prediction.Finally,the validity of the model should be verified using the virtual forcasting method and the probabilistic results can be get based on the historical relative errorsFurtherly, by making use of VC++.net programming language and SQL Server 2005 database, the system of traffic analysis and forecasting is implemented. Applying of the method to all the cells of the city of Guangzhou gives an efficient prediction.The forecasting error is within 10% through experimental verification.