Inference Of Anchor Links Between Multiple Social Networks

With the rapid development of Internet technology,online social networks have become an important channel for amusing,making friends,getting information and sharing information.In order to enjoy various excellent intelligent network services and meet personalized and diverse needs,people are usually involved in multiple online social networks simultaneously.Multi network anchor link inference,which is to infer the correlations between different accounts of the same person in different social platforms,is a prerequisite for many cross network applications,such as the cross-network social link recommendation,cross-network information fusion.The research on anchor link inference between multi networks is very important.In this paper,we clearly describe the research content of multi network anchor link inference problem,fully analyze the current research status and summarize the shortcomings of user document similarity based comparisons and network topology based analyses.Then we propose a MALI(meta path based multi network anchor link inference)to better solve the anchor inference problem.The main contributions of this article are as follows:1.The anchor link inference algorithm based on network topology is very sensitive to the noise and missing data problems,a PU link prediction model is proposed to build structure-rich probabilistic networks to against the noises.Further more,structure-rich networks can provide more abundant social information to improve the performance.2.Complex structure of partially aligned multi heterogeneous networks causes the diffi-culty in extracting useful cross-network features.In this paper,9 kinds of intra network social meta path are proposed and based on the small part of known anchor link,in-ter network social meta path could be established to extract powerful cross-network feature vector,which can well model the complex relationships between user nodes in heterogeneous networks.3.Considering that in order to distinguish with his/her friends in the social circles,each person usually tend to have a different user name.After establishing the user ranking list,we propose to incorporate username with the stable match algorithm to further improve the results.Extensive experiments on two real-world partially aligned heterogeneous networks show that our proposed MALI can solve the multi-network anchor link inference problem very well.Finally,this paper applies the results of multi network anchor link inference to two practical problems,Cross network social recommendation and identification of potential users.