It has been known for a long time that citation networks are always highly
clustered, such as the existences of abundant triangles and high clustering
coefficient. In a growth model, one typical way to produce clustering is using
the trid formation mechanism. However, we find that this mechanism fails to
generate enough triangles in a real-world citation network. By analyzing the
network, it is found that one paper always cites papers that are already highly
connected. We point out that the highly connected papers may refer to similar
research topic and one subsequent paper tends to cite all of them. Based on
this assumption, we propose a growth model for citation networks in which a new
paper i firstly attaches to one relevant paper j and then with a probability
links those papers in the same clique to which j belongs. We compare our model
to two real-world citation networks - one on a special research area and the
other on multidisciplinary sciences. Results show that for the two networks the
in-degree distributions are matched and the clustering features, i.e., the
number of triangles and the average clustering coefficient, are well
reproduced.
Description
[1104.4209] Modeling the clustering in citation networks
%0 Generic
%1 Ren2011
%A Ren, Fu-Xin
%A Cheng, Xue-Qi
%A Shen, Hua-Wei
%D 2011
%K citation clustering info20 network
%T Modeling the clustering in citation networks
%U http://arxiv.org/abs/1104.4209
%X It has been known for a long time that citation networks are always highly
clustered, such as the existences of abundant triangles and high clustering
coefficient. In a growth model, one typical way to produce clustering is using
the trid formation mechanism. However, we find that this mechanism fails to
generate enough triangles in a real-world citation network. By analyzing the
network, it is found that one paper always cites papers that are already highly
connected. We point out that the highly connected papers may refer to similar
research topic and one subsequent paper tends to cite all of them. Based on
this assumption, we propose a growth model for citation networks in which a new
paper i firstly attaches to one relevant paper j and then with a probability
links those papers in the same clique to which j belongs. We compare our model
to two real-world citation networks - one on a special research area and the
other on multidisciplinary sciences. Results show that for the two networks the
in-degree distributions are matched and the clustering features, i.e., the
number of triangles and the average clustering coefficient, are well
reproduced.
@misc{Ren2011,
abstract = { It has been known for a long time that citation networks are always highly
clustered, such as the existences of abundant triangles and high clustering
coefficient. In a growth model, one typical way to produce clustering is using
the trid formation mechanism. However, we find that this mechanism fails to
generate enough triangles in a real-world citation network. By analyzing the
network, it is found that one paper always cites papers that are already highly
connected. We point out that the highly connected papers may refer to similar
research topic and one subsequent paper tends to cite all of them. Based on
this assumption, we propose a growth model for citation networks in which a new
paper i firstly attaches to one relevant paper j and then with a probability
links those papers in the same clique to which j belongs. We compare our model
to two real-world citation networks - one on a special research area and the
other on multidisciplinary sciences. Results show that for the two networks the
in-degree distributions are matched and the clustering features, i.e., the
number of triangles and the average clustering coefficient, are well
reproduced.
},
added-at = {2011-08-12T12:35:08.000+0200},
author = {Ren, Fu-Xin and Cheng, Xue-Qi and Shen, Hua-Wei},
biburl = {https://www.bibsonomy.org/bibtex/2d668e639ed78f4c7ec53eeba64d8ae2a/sdo},
description = {[1104.4209] Modeling the clustering in citation networks},
interhash = {2aab1505ce7da27402449873fb57b48e},
intrahash = {d668e639ed78f4c7ec53eeba64d8ae2a},
keywords = {citation clustering info20 network},
note = {cite arxiv:1104.4209},
timestamp = {2011-08-12T12:35:08.000+0200},
title = {Modeling the clustering in citation networks},
url = {http://arxiv.org/abs/1104.4209},
year = 2011
}