Random Walk-based Beneficial Collaborators Recommendation Exploiting Dynamic Research Interests and Academic Influence

Abstract

It is laborious for researchers to find proper collaborators who can provide researching guidance besides collaborating. Beneficial Collaborators (BCs), researchers who have a high academic level and relevant topics, can genuinely help researchers to enrich their research. Though many efforts have made to develop collaborator recommendation, most of existing works have mainly focused on recommending most possible collaborators with no intention to recommend specifically the BCs. In this paper, we propose the Beneficial Collaborator Recommendation (BCR) model that considers the dynamic research interest of researcher's and academic level of collaborators to recommend the BCs. First, we run the LDA model on the abstract of researchers' publications in each year for topic clustering. Second, we fix generated topic distribution matrix by a time function to fit interest dynamic transformation. Third, we compute the similarity between the collaboration candidate's feature matrix and the target researcher. Finally, we combine the similarity and influence in collaborators network to fix rank score and mine the candidates with high academic level and academic social impact. BCR generates the topN BCs recommendation. Extensive experiments on a dataset with citation network demonstrate that BCR performs better in terms of precision, recall, F1 score and the recommendation quality compared to baseline methods.