Dynamic Overlapping Community Detection Algorithm Based On Social Networks And Its Application

The research of dynamic overlapping semantic community is an important research content for people to analyze the development law of events,monitor the development trend of public opinion,and make personalized recommendations.The current dynamic community detection algorithm based on evolutionary clustering framework does not consider the influence of attributes on the community structure,which makes the community structure less accurate and cannot analyze the relationship between communities and topics;too much reliance on historical community structure leads to "error accumulation" of the algorithm,which makes the community detection results poor,and even produces The "avalanche effect".In view of the above problems,this paper conducts research on community detection algorithms for dynamic social networks,and the specific work is as follows:(1)Aiming at the existing evolutionary clustering dynamic community detection algorithm that relies too much on the initial community structure and fails to consider the influence of node attribute characteristics on the community structure,this thesis proposes a dynamic community detection algorithm based on the similarity between nodes of social networks.The algorithm uses the idea of evolutionary clustering to take into account the dynamic similarity of nodes at the previous moment into account the similarity between nodes at the current moment,constructs a weight network,and then uses Louvain’s algorithm to obtain the community structure at each moment.The influence of node attributes and dynamic factors on community structure is analyzed in the real data set,and compared with similar algorithms,the algorithm effectively improves the quality of the community structure.(2)The dynamic community detection algorithm based on social network has the situation that the theme is separated from the community,resulting in the low recognition of the theme of nodes within the community,and the law of community change cannot be analyzed according to the evolution of the theme.In this thesis,a dynamic community detection algorithm based on topic division is proposed,which uses node attribute features as themes,constructs weighted sub-networks by using the idea of evolutionary clustering,obtains topic sub-communities by using the improved Louvain algorithm,and merges communities with high similarity to realize the detection of dynamic overlapping semantic communities.In the real data sets DBLP and Facebook,it is verified that the proposed algorithm has better community quality than other traditional algorithms.By using information entropy and link density,the relationship between the theme and the community is calculated,and finally the law of the community of different themes changing over time is evolved.(3)Aiming at the application problem of dynamic overlapping community,this thesis designs and develops a film recommendation system based on dynamic overlapping community,which mainly develops from four aspects: demand analysis,system design framework,recommendation process and system implementation,The research value of the social network dynamic overlapping community detection algorithm in recommender system is proved by the system.