Classification of scientific networks using aggregated journal‐journal citation relations in the Journal Citation Reports

Abstract

I propose an approach to classifying scientific networks in terms of aggregated journal-journal citation relations of the ISI Journal Citation Reports using the affinity propagation method. This algorithm is applied to obtain the classification of SCI and SSCI journals by minimizing intracategory journal-journal (J-J) distances in the database, where distance between journals is calculated from the similarity of their annual citation patterns with a cutoff parameter, t, to restrain the maximal J-J distance. As demonstrated in the classification of SCI journals, classification of scientific networks with different resolution is possible by choosing proper values of t. Twenty journal categories in SCI are found to be stable despite a difference of an order of magnitude in t. In our classifications, the level of specificity of a category can be found by looking at its value of D RJ (the average distance of members of a category to its representative journal), and relatedness of category members is implied by the value of D J-J (the average J-J distance within a category). Our results are consistent with the ISI classification scheme, and the level of relatedness for most categories in our classification is higher than their counterpart in the ISI classification scheme.

A reference-based Hirschian similarity measure for journals

“…Self-reference is predominant in the great majority of journals and, in a significant number of cases, ''kindred'' citation links (e.g., among Series A and B of the same family of journals) further distort the picture. It seemed, therefore, reasonable to use rankbased measures instead of frequency-based ones [like the cosine-measure used by Chen (2008)]. The most obvious choices of rank-correlation measures, Kendall's tau and Spearman's foot rule, do not perform well for lists of different size with a large number of ties.…”

“…The analysis of the cross-journal citation matrix (mainly with factor analysis) led Leydesdorff (2004aLeydesdorff ( , b, 2006 to implement a spectacular visualization of journal-journal relations. Chen (2008) used a reference-based similarity concept to validate and improve the Journal Citation Reports classification scheme. The most recent literature (e.g., Janssens et al 2009;Zhang et al 2009) emphasize the advantages of hybrid techniques combining the benefits of semantic, bibliometric and other methods.…”

Visualizing the world's scientific publications

“…The accuracy of cocitation clustering has been improved by introducing reference pair proximities as a new variable (Boyack, Small, & Klavans, 2013). Chen (2008) has also applied the affinity propagation method to cluster scientific networks. Instead of clustering journals, an alternative method has been designed for constructing classification systems at the level of individual publications (Waltman & Eck, 2012).…”

Improving SCImago Journal & Country Rank (SJR) subject classification through reference analysis

“…According to Glänzel and Schubert (2003) ''classification of science into a disciplinary structure is at least as old as science itself''. However, the facet of human knowledge that Chen described as a ''complex and dynamic network'' (Chen 2008) likewise complicates the development of any reliable and representative disciplinary classification scheme that might allow to effectively delimit different subjects or disciplines configuring this highly complex network.…”