Abstract: Collaborator recommendation is an important application in research social networks. Recommending suitable collaborators for researchers is conducive to enhance academic cooperation and improve the collaborative exchanges among authors. To this end, CRISI method is proposed to effectively recommend peers with high intensity of cooperation and similar research interests. Our method considers the cooperation intensity (structure) between authors, the similarity of research interest (attribute), as well as the closeness of the community formed by the authors to be recommended. Specifically, firstly, an attribute graph of the author’s cooperative relationship is constructed based on the relationship between the author and the literature. Then, the author’s cooperation intensity and research interest similarity are calculated and the dual-weighted network is constructed accordingly. Thirdly, the author nodes with high influence and strong cooperation intensity are detected as seed. Finally, a fractional k-core community search method is designed to find a community that has a close working relationship with the author to be recommended. The experimental results show that 
CRISI method can achieve significant performance improvement over the existing methods.

Key words: cooperation strength, research interest similarity, attribute graph, fractional k-core, community search