The Design And Implementation Of User Matching Across Social Networks

Nowadays,a great variety of online social networks have become an important part of people’s lives.People share and obtain information on social platforms such as Facebook and Weibo.As different social networks present different functionalities,there will be some difference in users’behavior data.If we can match users’ account on multiple social networks,it is meaningful in building complete user profiles and maintaining cyber security.Social network user matching problem aims to identify all virtual accounts referring to one same person in multiple social networks.Social network user matching provides a viable solution for consolidating data from multiple sources,helps many social network applications,such as friend recommendation and content customization,thus improving user stickiness and increasing website revenue.Therefore,how to accurately identify the same user behind different accounts has become a hot topic of research.This thesis presents a two-stage social network user matching algorithm that can eliminate its dependence on seed users and iteratively match users across the social networks stage by stage.This thesis proposes a two-stage algorithm.The first stage is the initial user matching algorithm based on the username,utilizing usernames to preferentially match a small number of users;the second stage is the global user matching algorithm based on multi-dimensional features,and the algorithm starts from the matched initial users of the first phase,and uses breadth-first search strategy to select users to be matched,and uses the matching model to integrate feature similarities of username,text and network structure,then calculate the users’ matching probability and generate new matched users.The new matched users will be used to retrain the matching model,iteratively match the remaining users in the network.Besides,this thesis proposes a community partition based social network user matching algorithm,which includes four steps:community partition,community matching,parallel matching and global matching.The community partition based social network user matching algorithm matches users in parallel on the community level,thereby improving matching efficiency for large-scale networks.The experimental results show that social network user matching methods designed in this thesis are superior to the comparison algorithm in the performance,robustness and efficiency of matching.