A Multi-Level Analysis of Collaborations in Computer Science
MetadataShow full item record
Working in collaboration is common in today’s highly connected scientific community. By collaborating, researchers can solve challenging multi-disciplinary problems, increase knowledge dissemination as well as productivity. These and other advantages motivate the study of the collaboration patterns of researchers. Such patterns can be observed directly in networks of manuscript co-authorship. In such a network, nodes represent authors and the links between them indicate that they have co-authored a paper. Several researchers constructed and studied large-scale networks representing collaborations in Mathematics, Biology, Physics, and Neuroscience. Most studies have performed bibliometric analysis of scientific publications, evaluated and ranked scholars on their research performances, and studied structural characteristics of the collaboration networks. Certain studies on collaboration networks are focused to specific geographical regions (i.e., country or countries). Studies on longitudinal analysis of the collaboration networks have helped to understand the publication trends of researchers. Most studies have analyzed collaboration networks of either authors or institutions or country. To best of our knowledge there is no study that has analyzed collaboration networks on all the three levels. The increase in international collaboration is not only a trend of the 21st century, but one that has been noted in bibliometric studies. However, very few studies have examined this collaboration activity. Also, there is very little knowledge and understanding about the role and the nature of geographical proximity towards scientific collaborations. In this dissertation, we analyze the collaboration networks at several levels (i.e., authors, institutions, and countries) in the field of Computer Science. We perform longitudinal analysis on publication trends and investigate collaboration patterns based on various geographical factors, such as distance and location. We investigate authors’ affiliation trends and their average productivity. We investigate, if the size (i.e., number of authors) and subject diversity of a institute play any role towards average productivity of that institute. We also analyze if there is any correlation between scientific size of a country and the GDP of that nation. We then rank authors, institutions and countries to list the top collaborators and also rank authors, institutions and countries based on network metrics. Last, using visualization techniques we show how authors and institutions are distributed globally. The results indicate that co-authorship networks in Computer Science have network properties similar to real-world networks and can be categorized as a scale-free network. The longitudinal analysis on the publication trends depicts a shift in the trend on number of authors in a research publication. Our findings show that geographical proximity plays a vital role in the collaborations patterns of authors. We observed, a growth in the trend for international collaborations for institutions. We found that the scientific size of a country is correlated to the GDP of that nation.