Peculiarities of forecasting the level of web traffic in general purpose computer networks




computer network, recurrent neural network, wavelet, Hurst parameter, time series, web traffic


The article studies the process of predicting the level of web traffic in computer networks. The possibility of predicting web traffic on different time scales using recurrent neural networks using the wavelet schedule of the original time series is considered. The wavelet transform decomposes the data so that the underlying temporal structures of the original time series become clearly visible. The individual wavelet coefficients are predicted, then recombined to obtain the final prediction. It is shown that the average number of bytes transferred in one hour demonstrates predictability when using this method. Further justification of the model parameters is carried out with the help of additional experiments and analysis of web traffic data.


Almeida V., Bestravros A., Crovella M. and de Oliveira A. Characterizing reference locality in the WWW / V. Almeida, A. Bestravros, M. Crovella and A. de Oliveira // DIS '96: Proceedings of the fourth international conference on on Parallel and distributed information systems (December 1996) / – IEEE Computer Society, 1996. – Р. 92-107.

Aussem A., Murtagh F. A neuro-wavelet strategy for Web traffic forecasting / A. Aussem, F. Murtagh // Research in Official Statistics. – 1998. – № 1. – Р. 65-87.

Aussem A., Murtagh F. Web traffic demand forecasting using wavelet-based multiscale decomposition / A. Aussem, F. Murtagh // International Journal of Intelligent Systems. – 2001. – Vol. 16, Iss. 2. – Р. 215-236.

Bruce, A. and Gao, H.-Y. S+Wavelets User’s Manual, Version 1.0 / A. Bruce and H.-Y. Gao. – Seattle, WA:StatSci Division, MathSoft Inc., 1994.

Crovella M.E. and Bestavros A. Self-similarity in world wide web traffic evidence and possible causes / M.E. Crovella and A. Bestravros // SIGMETRICS '96: Proceedings of the 1996 ACM SIGMETRICS international conference on Measurement and modeling of computer systems (May 1996) / – Association for Computing Machinery, 1996. – Р. 160-169.

Daubechies I. Ten Lectures on Wavelets / I. Daubechies. – Society for Industrial and Applied Mathematics, Philadelphia, 1992. – 343 p.

Holschneider M. and Tchamitchian P. Les ondelettes en 1989, ed. PG Lemarie / M. Holschneider and P. Tchamitchian. – Berlin: Springer-Verlag, 1990. – 102 p.

Leland W. and Wilson D.V. High time-resolution measurements and analysis of LAN traffic: implications for LAN interconnections / W. Leland and D.V. Wilson. – IEEE Infocomm`91, 1991. – 12 p.

Lin T., Horne B.G., Tino P. and Giles C.L. Learning long-term dependencies in NARX recurrent neural networks / T. Lin, B.G. Horne, P. Tino and C.L // IEEE Transactions on Neural Networks, 1996. – Vol. 7, № 6. – P. 1329-1338.

Starck J.-L. and Bijaoui A. Filtering and deconvolution by the wavelet transform / J.-L. Starck and A. Bijaoui // Signal Processing, 1994. – Vol. 35, Iss. 3. – P. 195-211.

Murtagh F., Starck J.-L. and Bijaoui A. Multiresolution in astronomical image processing: a general framework / F. Murtagh, J.-L. Starck and A. Bijaoui // International Journal of Imaging Systems and Technology, 1995. – Vol. 6, Iss. 4. – P. 332-338.

Starck J.-L., Murtagh F. and Bijaoui A. Multiresolution support applied to image filtering and deconvolution / J.-L. Starck, F. Murtagh and A. Bijaoui // Graphical Models and Image Processing, 1995. – Vol. 57, Iss. 5. – P. 420-431.

Wan E.A. Finite Impulse Response Neural Networks with Applications in Time Series Prediction: Ph.D. Thesis / E.A. Wan; Department of Electrical Engineering, Stanford University. – 1993. – 153 p.

Willinger W., Taqqu M., Leland W.E. and Wilson D. Self-similarity in high-speed packet traffic: Analysis and modeling of Ethernet traffic measurements / W. Willinger, M. Taqqu, W.E. Leland and D. Wilson // Statistical Science, 1995. – Vol. 10, № 1. – P. 67-85.