Community Detection Across Multiple Social Networks Based On Overlapping Users

As an important part of OSN related research,community detection has received extensive attention from researchers.Community is a reflection of the relationships among users in OSN,and contains information which is very valuable.Meanwhile,the diversification of user needs leads to the emergence of different types of OSNs,among which information is different and complementary to each other.Due to the large data flows among different social networks,many research works,such as public opinion analysis,information retrieval,and content recommendation,rely on community detection across multiple networks.However,the lack of explicit associations between different social networks and their prominent differences in main functions have made community detection across multiple social networks a new challenge in community detection area.In this paper,overlapping users across multiple social networks are defined as users who have active accounts in multiple social networks.As the bridge for information flow between different social networks,overlapping users widely exist in each social network.To solve the aforementioned problem,this paper proposes a novel community detection method across multiple social networks based on the overlapping users.First,overlapping users are mined out of multiple social networks and filtered according to specific rules.Based on the result,a hybrid network model which containing overlapping users across multiple social networks is built.After that,a sub-community discovery algorithm?CMN_NMF is proposed to discover overlapping users sub-communities,inside which are all overlapping users who are closely related with each other.Then,we conduct community discovery work based on the overlapping users sub-communities in different social networks separately.Finally,the communities in different social networks that contain overlapping users from the same overlapping users subcommunity are merged to form a cross-social networks community.To validate the method proposed in this paper,we conduct experiments using the group data from two OSNs? Zhihu and Weibo as dataset.And we propose three algorithm evaluation indicators?textual similarity in sub-communities,textual similarity in cross-social networks community and implicit overlapping users discovery ratio.The results of our experiments show that the cross-social networks community detection algorithm proposed in this paper can better mine similar users in different social networks.Finally,we design and implement the prototype system of across multiple networks community which visualizes user relationship and user details in the form of web pages.