2017年4月20日
PAFit: an R Package for the Nonparametric Estimation of Preferential Attachment and Node Fitness in Temporal Complex Networks
 ,
 ,
 記述言語
 掲載種別
 研究論文（学術雑誌）
 DOI
 10.1038/srep32558
Many realworld systems are profitably described as complex networks that<br />
grow over time. Preferential attachment and node fitness are two simple growth<br />
mechanisms that not only explain certain structural properties commonly<br />
observed in realworld systems, but are also tied to a number of applications<br />
in modeling and inference. While there are statistical packages for estimating<br />
various parametric forms of the preferential attachment function, there is no<br />
existing package for a nonparametric estimation, which would allow finer<br />
inspections on the famous `richgetricher' phenomenon as well as provide clues<br />
to explain nonstandard structural properties observed in realworld networks.<br />
This paper introduces the R package PAFit, which implements statistical methods<br />
for estimating the preferential attachment function and node fitness<br />
nonparametrically, as well as a number of functions for generating complex<br />
networks from these two mechanisms. The main computational part of the package<br />
is implemented in C++ with OpenMP to ensure scalability to largescale<br />
networks. In this paper, we first introduce the main functionalities of PAFit<br />
through simulated examples, and then use the package to analyze a collaboration<br />
network between scientists in the field of complex networks. The results<br />
indicate the joint existence of `richgetricher' and `fitgetricher'<br />
phenomena in the collaboration network. The estimated attachment function is<br />
almost linear, which means that the probability an author develops a new<br />
collaboration is proportional to their current number of collaborators.<br />
Furthermore, the estimated fitnesses reveal many familiar names of the complex<br />
network field as top fittest scientists.
grow over time. Preferential attachment and node fitness are two simple growth<br />
mechanisms that not only explain certain structural properties commonly<br />
observed in realworld systems, but are also tied to a number of applications<br />
in modeling and inference. While there are statistical packages for estimating<br />
various parametric forms of the preferential attachment function, there is no<br />
existing package for a nonparametric estimation, which would allow finer<br />
inspections on the famous `richgetricher' phenomenon as well as provide clues<br />
to explain nonstandard structural properties observed in realworld networks.<br />
This paper introduces the R package PAFit, which implements statistical methods<br />
for estimating the preferential attachment function and node fitness<br />
nonparametrically, as well as a number of functions for generating complex<br />
networks from these two mechanisms. The main computational part of the package<br />
is implemented in C++ with OpenMP to ensure scalability to largescale<br />
networks. In this paper, we first introduce the main functionalities of PAFit<br />
through simulated examples, and then use the package to analyze a collaboration<br />
network between scientists in the field of complex networks. The results<br />
indicate the joint existence of `richgetricher' and `fitgetricher'<br />
phenomena in the collaboration network. The estimated attachment function is<br />
almost linear, which means that the probability an author develops a new<br />
collaboration is proportional to their current number of collaborators.<br />
Furthermore, the estimated fitnesses reveal many familiar names of the complex<br />
network field as top fittest scientists.
 リンク情報
 ID情報

 DOI : 10.1038/srep32558
 arXiv ID : arXiv:1704.06017