Research on Server Load Prediction Based on Wavelet Packet Theory

A server load forecast model is presented based on wavelet packet analysis in this paper. Firstly, the server load time series are decomposed and reconstructed by wavelet packet analysis based on the model in order to get many server signal branches with the same length of history series ; then the BP neural network prediction models are constructed respectively for these branches, and finally their predicted results were combined into final load value. Theory analysis and Experiments slum that the frequency of each signal branch after the original signal is decomposed by wavelet packet is relatively simple and the correlation becomes stronger, so they become easier to be forecasted. The proposed method is superior to traditional predicting approach.

[1]  Peter A. Dinda,et al.  Multi-resolution resource behavior queries using wavelets , 2001, Proceedings 10th IEEE International Symposium on High Performance Distributed Computing.

[2]  Hang-Hang Tong,et al.  Boosting feed-forward neural network for Internet traffic prediction , 2004, Proceedings of 2004 International Conference on Machine Learning and Cybernetics (IEEE Cat. No.04EX826).

[3]  Alireza Khotanzad,et al.  Multi-scale high-speed network traffic prediction using combination of neural networks , 2003, Proceedings of the International Joint Conference on Neural Networks, 2003..

[4]  Peter A. Dinda,et al.  Online Prediction of the Running Time of Tasks , 2001, Proceedings 10th IEEE International Symposium on High Performance Distributed Computing.

[5]  Hu Chang-zhen,et al.  Research on Combination Prediction of Web Traffic Based on Wavelets , 2006 .