Busy Telephone Traffic Modeling And Forecasting Based On Mobile Communication

With the sustainable development of China Mobile3G and the arrival of4G, theMobile users will continue to increase in the next three years, and the telephone trafficwill subsequently grow. Especially in major holidays, communication networks aremore susceptible to the influence of many factors such as the population flow、emergency, and it will be faced with the impact of high traffic. In order to improve theutilization of network resources, and prevent the network congestion, Mobilepolicy-makers must understand the operation of the current state of thecommunication network, and take appropriate measures to control and managepromptly. And the ability to accurate prediction for mobile busy traffic is the premiseof scientific operation and management of communication network. Therefore, trafficprediction has been an important subject that researchers are concerned and is one ofthe difficult research fields.The traditional time-series forecasting methods are generally applicable to theglobal smooth, uncorrelated residuals characteristics of time series data, but the actualtraffic data mostly are non-linear, irregular, non-stationary, so it will have certainlimitations when using the traditional prediction methods, thus will affect theaccuracy of the predictions. With the purpose of analyzing and predicting traffic timeseries more accurately, the research methods proposed in this article are as follows:1. From the perspective of many factors, this paper proposes a busymulti-factor-based traffic forecasting model. As research finding, the change trend ofmobile traffic not only has close relationship with historical traffic data, but alsoaffected by the trend of other factors, it is necessary to analyze the correlation oftraffic, and find out the key factors and put them as input variables, finally using theleast squares support vector machine which is optimized by modified particle swarmalgorithm to predict, thus we can achieve the purpose of accurate prediction. 2. For modern busy telephone traffic show non-stationary, self-similarity,multi-scale features, the paper presents a combination of traffic forecasting model thatfuse the wavelet transform and least squares support vector machine. Firstly deposethe traffic data that presents non-stationary characteristics using Mallat algorithm, getlow-frequency and high frequency components, and then conduct single branchreconstruction for low frequency and high frequency components, and predict thevarious components that are reconstructed using LS-SVM model respectively, finallycompose the traffic, experiments show that the combination forecasting model hashigher prediction accuracy and stability.3. In order to improve the prediction accuracy of busy mobile traffic, the paperproposes a least squares support vector machine prediction combination methods thatfuse the wavelet transform and particle swarm algorithm. Through comparison withthree kinds of prediction models, the combined forecasting model has higherprediction accuracy and strong stability.