Research On Overlapping Community Detection Algorithms Based On Cluster Ensemble And Graph Neural Network

With the rapid development of science and technology,the network form has undergone tremendous changes.How to quickly and accurately detect the overlapping community structure in a complex network plays a crucial role in analyzing the function of the network.At this stage,the existing overlapping community detection algorithms mostly detect overlapping communities from a single perspective such as nodes,edges,and subgraphs,and cannot comprehensively consider a variety of information for community detection;and although some overlapping community detection algorithms make full use of the network structure information,but with a large time and space overhead.In view of these limitations,this paper conducts an in-depth study.The specific research contents are as follows.Firstly,in view of the shortcomings of some algorithms that only focus on a single angle and cannot comprehensively consider multiple information for community detection,an overlapping community detection algorithm based on clustering ensemble is proposed.The algorithm first selects multiple non-overlapping community detection algorithms from different angles as the base clusterer,and divides the given network to obtain the partition set of non-overlapping communities.Then,by calculating the node-node similarity and node-community similarity and connecting them,the node vector representation of the fusion topology and community attribution is obtained.Finally,combined with the graph convolutional neural network,the extracted node features are trained to obtain the node-community connection matrix that can most directly reflect the network structure,and then the overlapping communities are obtained.Secondly,in order to solve the defect that the existing overlapping community detection algorithms have large time and space overhead in making full use of the structural information in the network,an overlapping community detection algorithm based on meta-community and attention mechanism is proposed.The algorithm constructs a meta-network based on the initial community and the relationship between the initial communities on the basis of obtaining the initial community division based on different clustering methods.The meta-network is divided twice by the optimized modularity,and the meta-community is obtained,and the node feature vector is obtained by mapping the node and the meta-community.Combined with the attention mechanism,the node feature vector is trained according to the influence of neighbor nodes,and overlapping communities are found.Finally,the two algorithms proposed in this paper are tested on real-world datasets and synthetic datasets,respectively,and the experimental results obtained by the algorithm and the existing overlapping community detection algorithms are compared and analyzed.The experimental results verify that the effectiveness of the algorithm proposed in this paper.