Prediction of Network Traffic Using Multiscale-Bilinear Recurrent Neural Network with Adaptive Learning
A prediction scheme for network traffics using Multiscale-Bilinear Recurrent Neural Network (M-BLRNN) with adaptive learning procedure is proposed and presented in this paper. The proposed predictor is a combination between M-BLRNN and adaptive learning procedure. In M-BLRNN, the wavelet transform is employed to decompose the original traffic signals into several simple traffic signals. In addition, the adaptive learning procedure is applied to improve the learning process at each resolution level in M-BLRNN with adaptive learning (M-BLRNN(AL)). Experiments and results on a Ethernet network traffic prediction problem show that the proposed M-BLRNN(AL) scheme converges faster than M-BLRNN. The prediction accuracies of M-BLRNN and M-BLRNN(AL) are very similar in terms of the normalized mean square error(NMSE).
Keywordsprediction time-series recurrent neural network
Unable to display preview. Download preview PDF.
- 3.Chun, Y., Chandra, K.: Time series models for the internet data traffic. In: 24th Conference on the Local Computer Networks, pp. 164–171 (1999)Google Scholar
- 4.Aimin, S., Sanqi, Li.: A predictability analysis of the network traffic. INFOCOM 1, 342–351 (2000)Google Scholar
- 7.Park, D.C., Tran, C.N., Lee, Y.: Multiscale BiLinear Recurrent Neural Networks and Their Application to the Long-Term Prediction of Network Traffic. In: Wang, J., Yi, Z., Żurada, J.M., Lu, B.-L., Yin, H. (eds.) ISNN 2006. LNCS, vol. 3973, pp. 196–201. Springer, Heidelberg (2006)CrossRefGoogle Scholar
- 12.Fowler, H.J., Leland, W.E.: Local Area Network Traffic Characteristics with Implications for Broadband Network Congestion Management. In: IEEE JSAC, pp. 1139–1149 (1991)Google Scholar