Global scientific collaboration: A social network analysis and data mining of the co-authorship networks

Co-authorship networks consist of nodes and numerous links indicating scientific collaboration of researchers. These networks could be studied through social networks analysis and data mining techniques. The focus of the article is twofold: the first objective is the analysis of the co-authorship networks of the top 60 countries that had the highest number of scientific publications in the world, and the second one is the discovery of collaboration patterns of highly cited papers of these countries. To do so, all scientific publications of the top 60 countries in all fields as well as their highly cited papers were included in the study period between 2011 and 2015. The research samples in the first part included 10,460,999 documents and in the second part encompassed 711,025 highly cited papers. Required data were extracted from web of science database. To analyse co-authorship networks, centrality indices and clustering coefficient were used. UCINET, Pajek, VOSviewer and BibExcel software were used to map co-authorship networks of the countries and to calculate indices. Finally, the discovery of collaboration patterns in highly cited papers is studied through association rules. The research data indicated that over 95% of documents has been produced by the top 60 countries. In addition, the USA, Germany, England, France and Spain launched the most co-authorship. Quantitatively, there have been the most powerful collaboration links between China and the USA, the USA and England, the USA and Germany, and the USA and Canada. The clustering data indicated that collaborations of the top countries of the world were in three main clusters. The Friedman test showed that there was a significant difference in the priorities of the countries for collaboration; and the USA, China, England, Germany, France, Japan and Italy are in the top priority for collaboration, respectively. The results of collaboration pattern in highly cited papers indicated that the USA participates in more than half of collaboration patterns for producing highly cited papers.