Research Of Telephone Traffic Forecasting Model Based On Neural Network

With the rapid development of China’s economy, the developing speed of China telecommunications industry enhanced and the number of mobile phone users grow rapidly. The demands for telecom service increase gradually. The competition among the major telecom companies is more and more fierce. In order to meet people needs for work and daily life, telecommunication companies want to increase market share in greatest degree, they need to make a major breakthrough in terms of innovation, service and technology and finally have their own characteristics. More and more complex network structure and emergency situations, in addition to other factors such as population movements, major holidays, national macro-control and so on can influence the change of telephone traffic and lead to its massive change. In order to avoid of communications network congestion lead by rapid telephone traffic change, the research of telephone traffic prediction is particular important. The accuracy of telephone traffic prediction has important guiding significance for network optimization, reconstruction, configuration, and performance evaluation. At the same time, it was essential to ensure network stability and security operations. The choice of traffic prediction methods is the base of improving the prediction accuracy. It becomes the focus of this paper.The characteristics of telephone traffic data includes nonlinearity, time-variant and self-similarity, a model that applies Elman neural network which is optimized by the combination of simulated annealing algorithm and particle swarm optimization algorithm. The model can make full use of the advantage of Elman neural network which has the capability of appro ximating nonlinear function without limitation, the advantage of particle swarm optimization algorithm which can search the optima in global space, the advantage of simulated annealing algorithm that can escape from local optima. The model can solve the problem of falling into local optima and utilize dynamic recurrent characteristic in Elman neural network. The simulation results show that the prediction accuracy of this mothered could be higher than others for telephone traffic prediction.The correlated factors are considered. A model named grey neural network which combines gray model and artificial neural network model is studied. To solve the problem that the parameters in grey neural network are difficult to determine, it is easy to fall into local optima lead to weak approximation and less prediction accuracy, the improved particle swarm algorithm which applies threshold of velocity is employed to search for the optimum. When the velocity of particle is less than the set threshold, the accelerated momentum is applied on the particle to reinitialize the particle velocity and position. The correlation analysis approach is applied on the telephone traffic data to find the correlated factors. F ive most important correlation factors data and telephone traffic data are considered as the input of the proposed model to train network and forecast future traffic data. The experimental results show that prediction accuracy of proposed model is high.