Research On Ensemble-Based Community Discovery Algorithms For Multilayer Networks

Community discovery aims to uncover the community structure embedded in complex networks,and is one of the important tasks in complex network analysis.However,most of the existing community discovery methods are aimed at single-layer network data,and less research has been conducted on the multi-layer network data that are widely available in the real world.In recent years,some scholars have conducted useful explorations and proposed corresponding algorithms for the community discovery problem of multilayer networks.Among them,the community discovery algorithms for multilayer networks based on ensemble learning have been widely used with the advantages of simple implementation and strong scalability.However,the existing methods neglect the heterogeneity between different layers of communities and the importance between different base community structures and community divisions in the ensemble process,making it difficult to obtain accurate community structures.To address these problems,this paper conducts a series of studies on the ensemble-based community discovery algorithms for multilayer networks.Specifically,the main research of this paper is as follows.(1)To address the problem that existing algorithms tend to ignore the heterogeneity of different communities and the importance of different layers and different community divisions in the ensemble process,a two-stage ensemble-based community discovery algorithm for multilayer networks is proposed.In the first stage,the base community divisions generated by each layer network is mainly ensembled locally by combining the information of the optimal community division structure in the base community division obtained from each other layer network respectively;in the second stage,the importance measures of the obtained local community division and each community structure are firstly measured,and then the final community division results are obtained by global weighted ensemble.Finally,the effectiveness and robustness of the proposed algorithm are verified by comparative experiments.(2)A bipartite graph ensemble-based community discovery algorithm for multilayer network is proposed to address the problems that existing algorithms tend to ignore the importance of different layers and different community divisions in the ensemble process and have low computational efficiency.Specifically,firstly,the importance of each layer in the multilayer network is calculated by calculating the correlation of nodes at different layers;secondly,the reliability of each community structure is calculated based on the information entropy,the weight of each community in the base community divisions is obtained by combining the layer importance and reliability of each community structure,the bipartite graph is constructed,and the final community division results are obtained using the bipartite graph community discovery algorithm.Finally,the effectiveness and efficiency of the proposed algorithm are verified by comparative experiments.The research results obtained in this paper not only enrich the research content of community discovery for multilayer networks,but also provide technical support for the analysis and mining of multilayer networks.