PACIS 2019 Proceedings


The overwhelming number of groups in Scientific Social Network (SSN) makes it a great burden for researchers to find their desirable groups. Therefore, it’s critical to automatically recommend groups to researchers according to their preferences. Moreover, recommending groups to researchers in SSN naturally belongs to a One- Class Collaborative Filtering (OCCF) problem. Therefore, in this paper, a Heterogeneous Network-based (HN) Group Recommendation (HNGR) method is proposed to recommend groups to researchers in SSN. Specifically, to consider both the direct and indirect relations between entities in SSN, HN is employed to conduct similarity calculation to support further negative instances extraction and Probabilistic Matrix Factorization (PMF) recommendation processes. Experiments were conducted on the real world CiteULike dataset. The experimental results demonstrate the effectiveness of the proposed HNGR method, and the superiority of HN for similarity calculation which indicates the necessity of considering the indirect connections between researchers and groups.