Testing a priority-based queue model with Linux command histories

TL;DR

This paper evaluates a priority-based queue model for human dynamics using Linux command histories, finding that the model does not fully capture the temporal dependencies in human activity patterns.

Contribution

The study introduces new tests based on mutual information and the arcsine law to assess queue models against real human activity data.

Findings

Most datasets fit the priority queue model in terms of distribution

The model fails to account for temporal dependencies in human activity

Human dynamics modeling should incorporate statistical dependencies over time

Abstract

We study human dynamics by analyzing Linux history files. The goodness-of-fit test shows that most of the collected datasets belong to the universality class suggested in the literature by a variable-length queueing process based on priority. In order to check the validity of this model, we design two tests based on mutual information between time intervals and a mathematical relationship known as the arcsine law. Since the previously suggested queueing process fails to pass these tests, the result suggests that the modelling of human dynamics should properly consider the statistical dependency in the temporal dimension.