Adaptive estimation of Haar wavelet transform parameters applied to fuzzy prediction of network traffic

In this paper, we propose an adaptive approach to estimate the energies of the wavelet and scale coefficients of the Haar wavelet transform used in the multifractal modeling of network traffic traces. Simulation results confirm that the estimates obtained for the modeling parameters in the wavelet domain are precise. In addition, we propose an equation to calculate the autocorrelation function of the underlying multifractal model in terms of these wavelet domain parameters. In order to enhance the prediction performance of network traffic traces, the autocorrelation function is used to update orthonormal basis functions in a fuzzy system. To validate the adaptive fuzzy prediction approach, simulations with real network traffic traces are carried out, showing that the proposed algorithm provides lower mean square errors than other algorithms in the literature.