A Nearest Neighbor Based Personal Rank Algorithm for Collaborator Recommendation

Nowadays, more and more scholars find their own research collaborators through social platforms for scientific research. Due to the information overload problem, how to recommend collaborators accurately has become an important issue. In addition, with the development of academic research, interdisciplinary studies are more and more common. Previous topic modeling methods and some other social friend recommendation algorithms are not suitable for the recommendation of scientific research collaborators. Inspired by random walk with restart (RWR) and PageRank approach, this paper provides a nearest neighbor based random walk algorithm (NNRW) to recommend collaborators. Compared to the fixed probability of walking in traditional random walk algorithm, NNRW achieves better performance because it incorporates the social network characteristics and the probability of walking depends on the historical cooperation of the target user.