Research On Overlapping Community Discovery Based On Local Information

In complex networks,the community phenomenon is a common phenomenon that includes the same characteristics among different individuals.The set of nodes that are closely connected between the nodes is represented as a community.The linking relationship between the nodes within the community structure is close,and the linking relationship between nodes in different community structures is relatively sparse.The nodes within the community structure often have some common characteristics.In a complex network,when a node belongs to multiple communities,there is overlap between the communities,forming overlapping nodes and overlapping communities.In real life,the scale of each complex network is often very large.Existing methods are very time-consuming when dealing with community discovery of large-scale networks.How to effectively discover overlapping communities is still a hot issue.This paper will study based on the neighbor information between individuals,and only consider local information for community discovery.It is more suitable for large-scale complex networks than based on network global information.Local expansion algorithms usually start with seeds or seed sets,and expand seeds by absorbing external domain nodes into communities until each node belongs to at least one community.The key step of the research is the selection and propagation method of seeds.How to choose the right seed and expand the community effectively remains a huge challenge.Therefore,this paper studies the existing local community discovery algorithms.The main contents include:(1)Aiming at the shortcomings of the uneven seed selection scheme of the existing local community discovery algorithm,a local extended community discovery algorithm based on weak group structure is proposed.The algorithm proposes a new indicator of node importance measurement,which not only considers the number of neighbors of a node,but also considers the closeness of links between neighbors.First select the unvisited and most important node and the most similar neighbor node.If the two selected nodes have common neighbors,the two nodes and their common neighbors are combined to form an initial seed weak group.it runs iteratively until all nodes have been visited.Secondly,based on the neighborhood information of the weak seed group,the similarity judgment is made,and if similar,they are merged until the weak seed group has been visited,forming the initial community structure.Finally,the community is optimized.If there is a node with no assigned community,it will be added to the most similar initial community,and the initial community structure with higher overlap will be merged.Experimental results show that the algorithm can effectively improve the quality of overlapping community division and identify overlapping nodes,and the algorithm has low time complexity.(2)Aiming at the problem that the existing local community discovery algorithm does not consider the link weight,this paper proposes a local community discovery algorithm based on node similarity.The algorithm first determines the weight of each edge in the network based on the similarity of the node and the node,and performs seed selection according to the edge weight;then,the local adaptation function of the edge weight is used to locally expand the seed node until each node has been visited;finally,the community is optimized to improve the final division results of the community.Experimental results show that the algorithm can effectively improve the quality of community division and can effectively reduce the time complexity.