Linear Recession Analysis In GSM Network Traffic And Load Prediction

Mobile networks are developing rapidly in China in recent years. The total number of mobile network subscribers and growth has already exceeded the fix line networks. One of the impact is that the designed network capability may be hehind the real traffic load when peak time arrives due to the fast traffic increasing, and bring damages to the network safety. It’s a common challlenge for almost every operator to have a better accuracy in building network capapcity which matches better to the future growth of subscriber amount. What’s more, the change in traffic volume also means changing marketing strategies and network architects for the operators. This leads to an urgent demand for more precise and automatication enabled prediction techniques.This article firstly gives a general introduction of the GSM network, on the technology and development path. Then explained an traffic predicition solution and show a case in which the solution is applied.At the mean time, the arther tries to combine the traffic prediction with geographic grid techniques. This is because some network elements like MSCs and BSCs are the hubs of other network elements like BTS. And the topology and homing relationships are varying along the time frequently. And set up the grid also helps to give references when to allocate the traffic in planning.Work described in this articles are summarized below:Introduce the mobile communication network industry background, and the research subject:traffic prediction, set the object; analyze the traffic characteristics of several frequently involved network elements. Explain the basic concepts and technologies of GSM netwok and statistic analysis theory.Use the data wharehouse and unary linear recession analysis to analyze the history traffic data. Set up models of the traffics of the peak times like holidays, spring festivals, build the prediction solution. Disperse the predicted value into grids and then by re-aggregating the traffic, the load of the network elements like BSC and MSC and even the processing unit can be predicted.by implementation of the solution designed above, get the predicted data in real case. Set up pro-active plan for the network in the real case. Monitor and evaluate the results and performance.