A Novel Probability Distribution for Modeling Internet Traffic and its Parameter Estimation

Abstract

A new probability distribution following from Lavalette's law has been proposed for modeling wireless Internet access session statistics collected in a public nationwide Wi-Fi network. We have derived maximum likelihood estimators for this distribution and found them for three data sets of session duration and traffic volume measurements. Goodness of all fits is nearly perfect. The introduced model outperforms the Pareto, truncated Pareto and modified truncated Pareto distributions. Its flexibility makes it a good candidate for describing both long- tailed and non-long-tailed data.