The Research Of Multiple Factors Inlfuence Of Traffic Forecasting Model In Wireless Communication

In recent years, along with the rapid development of mobile communicationnetwork scale, and which present a rising number of users, the increasingly complexphenomenon of the network structure directly lead to the rising of the telephonetraffic, especially in the traditional holidays, traffic will be a sharp surge tendency,which makes the communication network congestion and affect the communicationquality. Therefore, for the mobile operators, knowing the change trend of the traffic inadvance can improve its market competitiveness. Traffic demand forecasting iscrucial to network planning and configuration for mobile network operators. Theaccuracy of the predict results decide the future development of the enterprise.The traditional forecast method of traffic is based on historical traffic to predictthe future traffic, the commonly used prediction models are ARMA model, neuralnetwork and support vector machine, etc. In wireless communication networks, trafficis actually influenced by many factors, not only related to the historical traffic, butalso related to short messages, the number of users when busy, the number of userswhen boot, GPRS upward flow and GPRS downward flow, etc. The traditionalprediction model is more suitable for single factor, to predict the traffic influenced bymany factors will affect the precision of prediction, so in order to more accuratelypredict the time series of traffic, this paper puts forward the traffic forecast model,which is applicable to many factors, the corresponding research contents as follows:1. According to the characteristic of mobile telephone traffic data and its changefeatures during the May Day, puts forward the multi-factors grey model—the trafficprediction model support vector machine compensation. Choose the data related totraffic, and find the factors which are more closer correlation of traffic by greycorrelation analysis method. First, with the multi-factors grey model to forecast thetraffic, then further get the residual error sequence, and use the support vector to establish residual error prediction model, thus to realize residual error compensation.The simulation results show that the algorithm need less number of samples and theprediction accuracy is high.2. In order to further improve the prediction accuracy of traffic, puts forwardan improved grey traffic prediction model which supports vector machinecompensation by improve the kernel function of support vector machine. The kernelfunction affects the prediction accuracy. The wavelet kernel function adopted in thispaper has good generalization effect. The simulation result shows that the predictionhas a higher speed and accuracy.