Study On Science Structure Based On Journal Clustering

The study of science structure is an import topic among the fields of science of science, management of science and technology, information science etc. Based on cross-citation analysis, clustering analysis and visualization technology, the whole knowledge structure and the information flow characteristics of natural science, social science and arts & humanity are quantitatively analyzed and presented.Firstly, on the basis of the cross-citation of "paper to paper", a huge cross-citation matrix of some 10,000 journals covered by Web of Science (2002-2006) is constructed. The analyses of science structure is based on this cross-citation matrix. A well-defined set of structure indicators including "strong links", "citation entropy" is proposed and designed for measuring the information flow characteristics, which provide some new methods and visual angles for better understanding of the internal structure of science. Based on different indicators, the degree of isolation, centrality, distribution of citation links of individual journal and subject in the communication network are analyzed and described. Some interesting characteristics of science structure are detected. For instance, there is clear divergence between journals having more "strong links" and those with "high-entropy"; subjects in social science and arts & humanity tend to distribute their links to a far-ranging scope, while subjects within natural science have more strong links.Secondly, a citation-based and a hybrid citation-text-based method are used to cluster 8305 journals and ISI 250 subject categories, respectively results in 22,15 and 7 journal clustering schemes, and 7 subject clusters. Some existing "expert" journal classification systems have been compared with the clustering structure, where the "coincidence" of the two schemes is defined as "subject cores", and the "difference" is "subject peripheries". All the subject cores form the nodes of science structure network, presenting the divergence of knowledge and the diversity of subjects. The subject peripheries present the trend of subject splitting, inter-crossing, convergence, method transplantation, and the interaction between theoretic subjects and applied subjects. For instance, "Neuroscience & behaviour" tend to split into two clusters, respectively with natural science and social science research preference. "Psychology/Psychiatry" in ESI tend to split into cognitive psychology and medical psychology clusters, while "Geosciences" and "Space science" tend to integrate into one cluster. Some theoretical subjects including mathematics have strong affinities with other subjects based on the "theory-application" relationship.Thirdly, based on citation analysis and visualization techniques, a global map of information flow characteristics in science, social science and art & humanities is generated. The information flow among different subjects reveals the intersection, convergence and unification trend of science and social science. The highly integration and unification of science is based on the highly specialities and differentiation. The theories of the unification trend of science proposed by Karl Marx and Qian Xuesen have been proved from the perspective of quantitative analysis. In the whole unification trend of science, "engineering", which is closely related to technological sciences serves as import communication roles. Furthermore, using some information exchange analysis between different subjects, for instance, between "sound" study and "psychology" study, some new growing points and underlying emergence of new inter-disciplines could be detected and revealed.