Study On The Approaches Of Semantic Overlapping Communities Detection

With the development of the network technology,the online social networks,such as Facebook,Twitter,have become the indivisible social channels.To enrich people's web community life,various social network sites launched the "Community Choice" and "Circle of Friends" services.Thus the community detection and recommendation algorithms are emerged,becoming the focus on social network data mining.Recently,with the context becoming the major carriers on social communication,the community detection orient to context analysis has become the new direction from the traditional community detection research field.The semantic community detection is mainly discussed in the dissertation.The major goal is to detect the communities with the tight semantic and topological relationship,furthermore find the overlapping nodes based on the detected communities,by the correlation analysis of semantic and topology of the nodes in the network.The research work can be carried out in the follows aspects: multivariate correlation research,potential research,local semantic community detection,semantic propagation research and semantic sampling research.Aiming at the problem that general social network community detection algorithm could only detect community with sole relationship which couldn't reflect the semantics similarity of real social community,which contained multivariate semantics topic,proposed the SSN(Semantics Social Network)detection algorithm based on topic comprehensive factor analysis.This algorithm defined the multivariate semantics information as topic,took multivariate TCF(Topic Comprehensive Factor)as the measurement of topics,the difference of topic density as assembly direction,established the initial community as first;Secondly,established the cost function with the goal of minimizing the semantics similarity inside the communities and maximizing the semantics similarity between the communities,when some nodes change the community;Finally,established the SAOP(Simulated Annealing Optimization Policy)based on the initial community and cost function,which could optimize the initial community globally,achieved the semantics community detection with multivariate.For the potential research,an overlapping community structure detecting method in semantic social network is proposed base on the semantic data field.Firstly,the algorithm utilizes the Gibss sampling method to establish the quantization mapping by which semantic information in nodes can be changed into the semantic space,with the LDA(Latent Dirichlet Allocation)as the semantic model;Secondly,establishes the semantic data field model,using the semantic coordinates and link relationships of nodes.Thirdly,proposes an improved Randwalk strategy of overlapping community structure detecting algorithm in SSN,with the semantic relation effort and the semantic potential of nodes as parameter.For the local semantic community detection,an overlapping community structure detecting method in semantic social network is proposed base on the Local Semantic Cluster(LSC).Firstly,the algorithm utilizes the Gibss sampling method to establish the quantization mapping by which semantic information in nodes can be changed into the semantic space,with the LDA(Latent Dirichlet Allocation)as the semantic model;Secondly,establishes the similarity matrix of SSN,with the relative entropy of semantic coordinate as the measurement of similarity between nodes;Thirdly,according to the character of local small-word in social network,proposes the S-fitness model which is the local community structure of SSN,establishing the LSC method by the S-fitness model.For the semantic propagation research,an overlapping community structure detecting method in semantic social network is proposed based on label propagation.Firstly,the algorithm utilizes the Gibss sampling method to establish the quantization mapping by which semantic information in nodes can be changed into the semantic space,with the LDA(Latent Dirichlet Allocation)as the semantic model;Secondly,proposes the principal component SCNP model which could measure the propinquity between nodes and the semantic impact model.Thirdly,proposes an improved label propagation algorithm,with SCNP as the weight of propagation and SI as the parameter of threshold.For the semantic sampling research,an overlapping community structure detecting algorithm in semantic social network base on the block field is proposed.Firstly,taking the LDA model as the semantic information model,the algorithm establish the BAT(Block-Author-Topic)semantic sampling model,with the sampling node as the core node.Secondly establish the measurement of the semantic cohesion of the block field,depending on the analysis of SSN,to achieve the measurement of semantic information.Finally,improve the label propagation algorithm for overlapping community detection,with the semantic cohesion as input.