Research On Mining Community From Multi-relational Social Networks

With the rapid development of Web2.0, there are more and more social networking sites. Social network analysis (SNA) has attracted more and more attention of researchers. Community mining is one of the major directions in social network analysis. Most of the existing methods on community mining assume that there is only one kind of relation in the social network, and moreover, the mining results do not fully meet the user's real need. However, in reality, there exist multiple, heterogeneous social networks, each representing a particular kind of relationship, and each kind of relationship may play a distinct role in a particular task. This thesis will be from the following three aspects:to reduce the noise data, to improve the efficiency of the algorithm and to utilize semantic information for community mining in social network. This research provides a method to solve these problems above. Specific research content mainly includes the following aspects:Firstly, this thesis proposes a multi-relational community mining algorithm based on correlation analysis (MCMABCA) for solving the impact of noise data. The method transforms community mining in multi-relational social network into the relationship of selection and extraction and reduces redundant relationship based on the correlation analysis. With the obtained combination relations combining user's query, it may mine community structure in multi-relational social network which meets users' need.Secondly, this thesis puts forward a multi-relational community algorithm based on ranking for solving algorithm complexity. With the user's query, each kind of relationship may play a distinct role. According to choosing a relationship set with a higher ranking of importance degree, this method will discovery community structure.Thirdly, sometimes, users may not get the mining results wanted only from the graph topology. Considering the semantic information in social network, the paper proposes a method of discovering semantic community. It firstly defines and constructs a community-entity-data model (CED model), and address the problem of semantic community discovery by adapting Gibbs sampling algorithm to the model.Finally, this thesis designs and achieves the experiments. Compared with traditional community mining methods, the experiments verify the correctness and effectiveness of above proposed algorithms or methods.