Approximation of Internet traffic in wavelet domain

Abstract

The paper elaborates the complexity and importance of modeling Internet traffic. In this sense, having in mind the nature of Internet traffic as well as actual information in this sphere, it was emphasized that there is no universally applicable model. This standpoint has initiated the idea of examining the possibility of approximation internet traffic. Because of the rather suitable property, namely that is not necessary to provide stationarity of the input signal, traffic analysis in this paper was performed in wavelet domain, using different types of wavelets, with the aim of determining the convenience of certain wavelets for the analysis of Internet traffic. It was concluded that all examined wavelets have generated approximately similar statistical results, which depend on the selected level of approximation and analyzed traffic curve. For different curves at the same level of approximation, statistical values have been obtained that do not differ significantly from the statistics of the corresponding original signals. It was also found that the discrete approximation of the symlet wavelet sym(2) provided the largest number of statistically accurate results, and in this regard proved to be the most suitable for the analysis of Internet traffic.