Anderson Paschoalon


Experienced software developer specializing in C/C++, embedded systems, and desktop applications. Master's degree holder and researcher with a focus on network technologies. Enthusiast Deep Learning and Game Modding.

Using BIC and AIC for Ethernet traffic model selection. Is it worth?

| View Publication

Abstract: In this work, we aim to evaluate how good are the information criteria AIC and BIC inferring which is the best stochastic process to describe Ethernet inter-packet times. Also, we check if there is a practical difference between using AIC or BIC. We use a set of stochastic distributions to represent inter-packet of a traffic trace and calculate AIC and BIC. To test the quality of BIC and AIC guesses, we define a cost function based on the comparison of significant stochastic properties for internet traffic modeling, such as correlation, fractal-level and mean. Then, we compare both results. In this short paper, we present just the results of a public free Skype-application packet capture, but we provide as reference further analyzes on different traffic traces. We conclude that for most cases AIC and BIC can guess right the best fitting according to the standards of Ethernet traffic modeling.

Published in: IX DCA/FEEC/University of Campinas (UNICAMP) Workshop (EADCA)

Date of Publication: 29 September 2016

Publisher: Departament of Computing Enginiering and Automation (DCA) – UNICAMP

BibTeX:

@misc{artigo2274:online,
author = {Anderson dos Santos Paschoalon , Christian Esteve Rothenberg},
title = {Using BIC and AIC for Ethernet traffic model selection. Is it worth?},
howpublished = {\url{https://www.fee.unicamp.br/sites/default/files/departamentos/dca/eadca/eadcax/trabalhos/artigo_22_Using_BIC_AID_Ethernet_Traffic_Anderson_Prof_Christian.pdf}},
month = {Octuber},
year = {2017},
note = {(Accessed on 07/13/2022)}
}