Research On Methods Of Overlapping Community Discovery And Fusion In Multidimensional Attribute Networks

Community discovery is an indispensable field in the study of social networks.Community discovery is mainly divided into overlapping community and non-overlapping community discovery.In real life,it is usually dominated by overlapping communities.Because of the multi-dimension of social network information,the diversity of users,the multilateralism of network structure and the efficiency of discovering communities,the grim situation faced by community discovery cannot be ignored.This paper mainly studies and analyzes the existing methods and related work from perspectives of social networks’ attribute information and users’ interaction behavior.The main research results are as follows.Aiming at the problem that the existing methods ignore the fusion of multi-semantic attributes of the network and variability of topological features under the spatial context perception,this paper proposes a new overlapping community detection method – SCEMR(Semantic Overlapping Community Detection with Embedded Multi-dimensional Relationships and Spatial Context).Firstly,on the basis of modeling users’ Microblogs in Microblog social networks by Latent Dirichlet Allocation for extracting the semantic topic information,are built to quantify the interest preference of user.Secondly,taking the network topology embedded with implicit semantic information as the core,the seed nodes are generated by leveraging entropy to fuse explicit information.Finally,under the spatial context,semantic overlapping communities are detected based on the Friend Probability in spatial groups and the Preference Matching Degree of users,which are utilized to deeply search the corresponding neighbor nodes in the process of detection.The effectiveness of this method has been verified on real data sets.The experimental results show that SCEMR can detect high-quality semantic overlapping communities in real microblog social networks.In order to solve the problems of weak blog information represented by topic probability and differences in user interest behavior characteristics.This paper proposes a community discovery method based on word vectors – SUIO(Semantic Overlapping Community Discovery Method Based on User Interests to Optimize Network Structure).Firstly,Word2 Vec is used to extract the semantic information of user behavior,and a user interest preference model is constructed to optimize the social network structure.Then,based on user interest preference,the traditional Page Rank algorithm is improved to optimize the seed nodes.Finally,local community discovery is carried out around the seed nodes.The rationality of SUIO was tested on three real datasets,and the results showed that semantic representation based on word vectors can more accurately model user interest preferences and improve the accuracy of semantic overlap community discovery.Aiming at the problem of community redundancy in the process of community detection,the current methods are mostly solved by the idea of community shared nodes,and do not fully utilize the network attribute information and user interest preferences.Therefore,this paper proposes an effective overlapping community merging method – EOCMM(An Effective Overlapping Community Merging Method Oriented to Multi-dimensional Attribute Social Networks).In EOCMM,we focus on two core problems which are how to improve the method of improving overlapping community detection in multidimensional attribute social networks and how to merge the detected communities with high redundancy degree.To solve the first problem,based on the network topology characteristics and user interest preference,the Node Domain index is proposed to improve and optimize the selection of seed nodes.To solve the second problem,we merge the communities with high redundancy degree by Semantic Community Overlapping Degree which is fused by user similarity,community similarity and community tightness.The experimental results show that this method can make the results of community detection and merging more reasonable and effective.