Scale-free download network for publications

TL;DR

This paper reports the first observation of scale-free, power-law behavior in publication download frequencies, using empirical data and modeling with the Barabasi-Albert network to explain the phenomenon.

Contribution

It introduces the first empirical evidence of scale-free download networks for publications and applies the Barabasi-Albert model for explanation.

Findings

Download frequency follows a power-law distribution with specific exponents.

Zipf-law and Tsallis entropy methods effectively fit the download data.

The Barabasi-Albert model explains the network's scale-free nature.

Abstract

The scale-free power-law behavior of the statistics of the download frequency of publications has been, for the first time, reported. The data of the download frequency of publications are taken from a well-constructed web page in the field of economic physics (http://www.unifr.ch/econophysics/). The Zipf-law analysis and the Tsallis entropy method were used to fit the download frequency. It was found that the power-law exponent of rank-ordered frequency distribution is $\gamma \sim 0.38 \pm 0.04$ which is consistent with the power-law exponent $\alpha \sim 3.37 \pm 0.45$ for the cumulated frequency distributions. Preferential attachment model of Barabasi and Albert network has been used to explain the download network.