Link Prediction in Co-authorship Network
Research Area: Digital Libraries Year: 2014
Type of Publication: Article Keywords: Link prediction, supervised learning, co-authorship network
  • Minh Le Nhat
Honours Year Project Report
Link prediction is one highlighted aspect of network analysis. Many popular applications include suggesting friends on Facebook and LinkedIn are using link prediction to feature its services. As a researcher, we are particularly interested in link prediction for scholar social network. In the research area, collaborations among authors often form a network of connections which defines the co-authorship network. It is proved that collaborations between authors/researchers often yield more fruitful results than individual researchers‟ performance. It comes to a sense that individual researcher might find himself want to collaborate with potential researchers. Predicting the evolution of co-authorship network thus plays a significant role in (1) analyzing the trend of the structure of scientific collaborations; (2) detecting potential research communities as well as their evolutions; (3) assessing the future influence of scientists; and (4) recommending companions, assistants, or colleagues for individual researchers. Realized the importance of link prediction, we are motivated to enhance its performance especially in co-authorship network. In this work, we propose new features that help to predict the collaborations among scholars, and analyze how these features work.
Digital version