Mobile Traffic Prediction Method Based On Echo State Network Model

With the development of communications industry, the quality of communication greatly depends on the allocation and assessment of channel capacity in mobile communications. Hence, as measure criteria of channel capacity, mobile traffic has drawn more and more attention. The accurate prediction of the traffic not only makes the channel efficiently utilized but also effectively prevents the channel chocking. Currently, the prediction of mobile traffic has become an important problem to most scholars.Aiming at this problem, this paper carries out the following research work.Firstly, mobile traffic prediction schemes based on ESN model was proposed after the analysis of mobile traffic data and the study of ESN (echo state network) theory. Matlab and C-language simulation platform are built to test the proposed prediction method. From the perspective of software simulation, the proposed methods demonstrate the good feasibility and effectiveness. Compared the simulation results, it is easy to see that the prediction accuracy of both two methods are basically the same, but C-language simulation shows higher computing efficiency.The proposed mobile traffic prediction schemes based on ESN model has higher prediction accuracy, but still can be further improved. What's more, ESN model is not suitable for solving the prediction problem of the multi-scale mobile traffic sequence. For this issue, the wavelet Analysis which is a classic method that handle prediction problem of multi-scale sequence is introduced, so Combination of Wavelet Analysis and ESN model prediction method is proposed.(1) Wavelet Echo State Network Model:the embedded type is used, in which wavelet neural became as basic information processing unit. This model is used to predict the noisy multi-scale sin sequence, and the results verified the validity of the model. The model is successfully applied to the prediction problem of the mobile traffic.(2) For the recent existing problem of boundary effects and the accumulation of error, Wavelet Decomposition Echo State Network based on cascade type, Anti-Boundary Effects Echo State Network is proposed. This model predicts three sequences in ESTSP 2008 meeting and the noisy multi-scale sin sequence. The results verified the validity of the model. Then model is successfully applied to the prediction problem of the mobile traffic.