Skip to main content


  • Original Paper
  • Open Access

Ranking in collaboration networks using a group based metric

Journal of the Brazilian Computer Society201117:41

  • Received: 23 February 2011
  • Accepted: 15 September 2011
  • Published:


Collaboration networks are social networks in which relationships represent some kind of professional collaboration. The study of collaboration networks can help identify individuals or groups that are important or influential within a given community. We start this work by characterizing the structural properties of the scientific collaboration network in the area of Computer Science. In particular, we consider the global network (all individuals) and the Brazilian network (individuals affiliated with Brazilian institutions) and establish a direct comparison between them. Our empirical results indicate that despite exhibiting features found in most social networks, these two networks also have some interesting differences. We then present a novel approach to rank individuals within a group in the network (as opposed to ranking all individuals) using solely their relationships. Intuitively, the importance assigned to an individual by our metric is proportional to the intensity of its relationship to the outside of the group. We use the proposed approach and other classical metrics to rank individuals of the Brazilian network and compare the results with the ranking of the Research Fellowship Program of CNPq (an agency of the Brazilian Ministry of Science and Technology). The direct comparison indicates the effectiveness of the proposed approach in identifying influential researchers, in particular when considering top ranked individuals. We then extend the proposed approach to rank small groups of individuals (as opposed to single individuals). We apply this and other classical metrics to rank graduate programs in Computer Science in Brazil and compare the results with the ranking of graduate programs provided by CAPES (an agency of the Brazilian Ministry of Education). Our results indicate that the proposed method can effectively identify influential groups such as well-established graduate programs in Brazil.


  • Collaboration networks
  • Network structure
  • Node ranking