Research On Overlapping Community Detection Based On High-impact Edge For Complex Networks

Many complex systems in the real world can be abstractly represented as network forms.With the improvement of the availability of large-scale network data and the indepth study of quantitative and qualitative characteristics of complex networks by scholars,large-scale complex networks have become a very significant Research topics,and the discovery of community structure in complex network data is one of the difficult points that have been focused on.In 2017,an overlapping community discovery algorithm named central nodes selection algorithm(CNS)based on gravity was proposed,which was based on selecting seed points around the center point of the community and then attracting other points around it.Then,in 2019,one celestial body was integrated into the other.The overlapping community discovery algorithm based on the selection of the center edge(CES)is a further improvement of the traditional CNS,but because the CES algorithm continues the selection process of the CNS algorithm’s seed nodes,it leads to a certain degree of quality evaluation of its nodes Bias,so it is easy to produce the possibility that the influence function produces inaccurate results in the description of the online community,which eventually leads to a situation where the division result is inconsistent with the real result.The situation is still not accurate.Based on the selection process and clustering process of CES seed nodes,this paper proposes an algorithm based on high-impact edges.The previous algorithm has been improved in the following three parts: Use neighbor node contact neighbor node connectivity(NNCD)to adjust the quality of the node to obtain a more reasonable seed node.NNCD is based on the number of cycles between nodes,which can make nodes with more cycles between nodes have larger points.Similarity between nodes;added sorting during the search of seed nodes to avoid incomplete problems with nodes with similar structures;using the improved link similarity(ILS)formula to divide non-center edges into The correct class In this class,this process redefines the selection process of non-central edges,so that the effect of community division is closer to the actual network.The experimental results on three real networks and three protein interaction networks show that the algorithm proposed in this paper is more effective than the CES algorithm.The CNS algorithm,CPM algorithm,LC algorithm and other algorithms have been further improved.