Collaborator Recommendation Based On Scholarly Collaboration Data

In recent years,with the rapid development of information technology,the scale of academic cooperation network developed greatly.Moreover,academic cooperation as an important way of communication in the academic field,more and more scholars have invested into academic cooperation.Through academic cooperation,scholars have also obtained many benefits.How does cooperation promote the progress of scholars? How to find the most valuable collaborators more quickly? Therefore,this paper aims at these two problems.Academic cooperation profit is quantified by social informatics method.We also propose a collaborator recommendation strategy based on academic cooperation data.Academic cooperation usually has a positive impact on scholar's productivity.How does cooperation promote scholars' influence in academic networks? We use the social informatics methods to quantify the analysis,mainly on the network structure and individual characteristics.At the same time,we introduce Bacon number theory,generalized friendship paradox and ternary closure theory to analyze the effectiveness of cooperation.In many cases,it is time-consuming 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 research topics,can genuinely help researchers to enrich their research.However,how can we find the BCs? In this paper,we propose BCR model.BCR runs a topic model on researcher's publications and associates three academic features: topic distribution of research interest,interest variation with time,and impact in collaboration network.First,we run the LDA model on researchers' publications for topic clustering.Then,we compute the similarity between the collaboration candidate's feature matrix and the target researcher.Third,we fix the generated topic distribution matrix by a time function.Finally,we fix the rank score by combining the similarity and influence.Extensive experiments on the citation network dataset show that,in comparison to relevant four baseline approaches,BCR performs better in terms of precision,recall,F1 score and the recommendation quality.