A Study of Wavelet Analysis and Data Extraction from Second-Order Self-Similar Time Series.

Abstract

"In the past decades, the self-similar processes and long-range dependence (LRD or long memory) have been applied to the study and modeling of many natural and man-made complex phenomena. These kinds of processes have been particularly attractive in the pursuit of optimal design and configuration of network communications. The published work of Leland et al. in 1993 and 1994 [1, 2] demonstrated that Ethernet traffic is statistically self-similar and that the commonly used models are unable to capture that fractal behavior, highlighting that a burstiness and LRD are present when �� > 0.5. Since then, researchers have been studying extensively long memory processes and their impact on network performance, for example, Karagiannis et al. stated that the identification of LRD is not trivial and that not all scenarios in modern networks present LRD characteristics, for example, traffic in the Internet backbone is more likely to be Poisson type instead of LRD [3]. Many researchers have also addressed their"