The Research Of Traffic Forecasting Model In Wireless Communication

With the continuous development of mobile communication industry for the pastfew years, the network structure is under extensive construction and becoming moreand more complex, the number of users is increasing and telephone traffic is alsorising. Especially, the telephone traffic will surge in the major holidays so that thewireless network is under the shock of high telephone traffic, which requires mobilecommunication operators to carry out planning and optimizing the network in time.Grasp the development trend of future telephone traffic is the basis and premise of thenetwork reconstruction and configuration. Therefore, the searching of telephonetraffic prediction model, which has important directive significance to themanagement and the operation of the mobile communication network, so to a certainextent it can guarantee security and stable operation of the network, and also canreduce the operation cost and allocate resources reasonable.Methods of traditional time-series forecasting require time-series with the globalstationary, uncorrelated residuals, but in fact, traffic data is nonlinear, irregular, andnon-stationary. In order to analyze and forecast traffic time series accurately, theresearch contents of this paper are just as follows:1. According to the characteristics of the traffic data, traffic prediction modelbased on particle swarm optimization (PSO) to improve the Elman neural network ispresented. PSO algorithm is used to optimize and train the dynamic recursion networkparameters and improve the generalization ability of the network. The experimentresults indicate that the method of the Elman Neural Network combining with ParticleSwarm Optimization algorithm can achieve a higher prediction accuracy and fasterconvergence speed.2. Through the statistical analysis of historical data of traffic, puts forward a kindof prediction mode based on the improved algorithm of PSO to optimize the leastsquares support vector regression (LS-SVR). The selection of hyper parameters ofLS-SVR is correct or not determines its prediction ability, improved particle swarmalgorithm realizes the optimization of parameters by the diversity measure whichkeeps the population activity. Results of traffic prediction show that the improvedPSO algorithm overcomes the disadvantage that basic particle swarm algorithm iseasy to fall into premature state, the algorithm also speed up the convergence, cansearch to the optimal solution quickly with high accuracy.