Research On Hybrid Friend Recommendation Method For Scientific Research Social Networks

Friend recommendation has taken an important position in the field of personalized recommendation,and has become research focus in both industrial and academic societies recently.Scientific social networks,as one of most typical application with Web 2.0 technology,has been popular in recent years,because it can meet researchers personalized social needs.In order to solve information overload problem and promote academic collaboration,we proposed a friend recommendation approach in scientific social networks domain.Therefore,it is of great theoretical and practical significance to consider the problem of friend recommendation for users in scientific social networks.Meanwhile,exist researches on friend recommendation mainly focus on public social networks,few researchers considered the context of scientific social networks.Moreover,many friend recommendation methods are based on the interaction behaviors between users and friends.In fact,there are plenty of social information in scientific social networks are not fully mined.Therefore,a novel hybrid method combining content-based filtering method and matrix factorization is proposed and presented as a prettier way to model the friend recommendation problem in this paper.Firstly,this method incorporates TF-IDF and sentiment analysis technology into traditional content-based filtering method.Then combines matrix factorization with users' similarity and social relationship,and generates friend recommendation list.Lastly,in order to verify the effectiveness of the proposed method,the experiments were conducted on the Science Net dataset.The experimental results show that our proposed method in this paper has achieved better results at the evaluation metrics.It is indicated that the proposed method which considering both users' similarity and social relationship can efficiently improve recommendation performance.