Prediction of server load based on wavelet-support vector regression-moving average

To improve the predication accuracy for server load, a novel predication method was proposed based on the integration of wavelet analysis and support vector regression. The server load time series, which is nonlinear and non-stationary, was decomposed and then, reconstructed into several branches by the wavelet method. Of these branches, the lowest scale high frequency signal was forecasted by moving average model, the others were predicted by support vector regression respectively and the final value was the combination of these predicted results. Theoretical analysis and experiment results show that wavelet analysis can decompose the original load series into several time series that have simpler frequency components and are easier to be forecasted; support vector regression has greater generation ability and guarantees global minima for given training data, it performs well for non-stationary time series prediction. So the method has higher predictive precision than traditional prediction approaches.