We analyse the large-scale structure of the journal citation network built from information contained in the Thomson-Reuters Journal Citation Reports. To this end, we take advantage of the network science paraphernalia and explore network properties like density, percolation robustness, average and largest node distances, reciprocity, incoming and outgoing degree distributions, as well as assortative mixing by node degrees. We discover that the journal citation network is a dense, robust, small, and reciprocal world. Furthermore, in and out node degree distributions display long-tails, with few vital journals and many trivial ones, and they are strongly positively correlated.
The present study is an interdisciplinary research integrating the fields of network science and bibliometrics. The field of network science -the holistic analysis of complex systems through the study of the structure of networks that wire their components -exploded in the last decade, boosted by the availability of large databases on the topology of various real networks, mainly the Web and biological networks. The network science approach has been successfully applied to analyse disparate types of networks, including technological, information, social, and biological networks (Brandes and Erlebach, 2005;Newman et al., 2006;Newman, 2010). Network analysis can be performed at different levels of aggregation:
• Node-level analysis. At this level, the goal is to measure the importance or centrality of a node within the network. Centrality here is not an intrinsic and permanent feature of the node but, instead, it is an extrinsic and fleeting property that depends on the interactions of the node with the other nodes in the network. Typical node centrality measures include degree, eigenvector, closeness and betweenness centrality.
• Group-level analysis. This investigation involves methods for defining and finding cohesive groups (clusters) of nodes in the network. The definition of cluster depends only on the topology of the network. Clusters are tightly knit sets of nodes with many edges inside the cluster and only a few edges between clusters. Two typical methods at this level of analysis are graph partitioning (where the number of clusters is fixed in advance) and community detection (in which the number of clusters is unspecified).
• Network-level analysis. The focus of this analysis is on properties of networks as a whole such as connectivity, mean and largest distances among nodes, distribution of node degrees, frequency of topological motifs, and assortative/disassortative mixing. It also includes the investigation on theoretical models explaining the generation of networks with certain features (e.g., random, small-world, and scale-free models).
Bibliometrics is an older field; it is a branch of information and library science that quantitatively investigates the process of publication of research achievements (Garfield, 1955;de Solla Price, 1965). Networks abound in bibliometrics; two important examples are citation networks of articles, journals or disciplines and collaboration networks of scholars. Other bibliometric networks are co-citation and co-reference networks of articles, journals or disciplines.
Collaboration networks have been largely studied using the network science approach (Newman, 2004;Barabási et al., 2002;Grossman, 2002;Moody, 2004;Radicchi et al., 2004;Franceschet, 2011). Journal citation networks have been investigated mainly at node-and group-levels. The investigation at the nodelevel concerns the proposal of eigenvector-based centrality measures for journals (Pinski and Narin, 1976;Bollen et al., 2006;West et al., 2010), the clustering of journal bibliometric indicators, including centrality measures, on the basis of the statistical correlation among them (Leydesdorff, 2009;Bollen et al., 2009), as well as the use of betweenness centrality as an interdisciplinary indicator for journals (Leydesdorff and Rafols, 2011). The group-level analysis of journal citation networks focuses on the detection, using different methods, of communities of journals, which correspond to fields of knowledge in the map of science (Leydesdorff, 2004;Rosvall and Bergstrom, 2008;Klanans and Boyack, 2009;Leydesdorff et al., 2010).
The investigation of journal citation networks at the network-level has been mainly focused on the study of the distribution of citations among papers and journals (Seglen, 1992;Redner, 1998;Stringer et al., 2008;Radicchi et al., 2008). The aim of the present investigation is to complement this investigation with additional large-scale structure properties of journal citation networks. More specifically, we focus on the following network properties: density of citation links, robustness with respect to the removal of nodes according to different percolation strategies, average and largest path lengths, topological motifs of reciprocity, incoming and outgoing degree distributions and their statistical correlations, as well as assortative mixing with respect to incoming and outgoing node degrees.
We considered all science and social science journals indexed in Thomson-Reuters Journal Citation Reports. We built a journal citation network in which the nodes are the selected journals and there is a directed edge from node A to node B if journal A published in 2008 a paper that cites a paper printed in journal B in the temporal window between 2003 and 2007. We only took into account the document types article and review. We considered the sub-network corresponding to the largest strongly connected component of the original network, which covers the large majority of the origin
This content is AI-processed based on open access ArXiv data.
