Research On Overlapping Community Discovery Method Based On Deep Clustering Fusion

Complex networks have good network transfer and network aggregation characteristics,can accurately model complex systems,and have become the focus of research in many fields such as social sciences,biomedicine,physics,and computer science.Community structure is a universal characteristic of complex networks.It is common in social networks,protein biological networks,transportation networks,and power networks.Studying the community structure can better understand the internal structure and data distribution of complex networks,which has good theoretical significance and practical value.Most of the community structures of complex networks have overlapping phenomena.As a link between different communities,it can more realistically reflect the real world.How to efficiently find and judge overlapping communities poses higher challenges for research.This paper explores and researches on the method of discovering overlapping communities in complex networks.Although existing research work has achieved certain results,there are still problems such as poor generality of the algorithm,low accuracy,and easy to fall into local optimum.In view of the above problems,this paper applies clustering fusion methods with high applicability,stability and scalability to the discovery of overlapping communities in complex networks.As the graph structure model is directly processed by the cluster fusion method,it will destroy the topological structure of the complex network and is susceptible to parameters.This paper improves the cluster fusion method and proposes the SOM-GCN algorithm based on deep cluster fusion for the overlapping community discovery of complex networks.Introduce self-organizing neural network SOM to improve the process of generating cluster members,retain the topological structure of the overlapping community of complex networks,and obtain high-quality cluster members while ensuring the integrity of the overlapping structure;use the map volume that can deeply dig the topology graph structure The product neural network GCN model further optimizes the consensus function,taking into account the structural characteristics of the graph nodes while taking into account the structural characteristics,greatly improving the efficiency of identifying overlapping communities.Experimental results show that the SOM-GCN algorithm proposed in this paper has higher accuracy,better stability,and is less affected by parameters compared with the classic algorithm.