Research On Community Discovery Algorithms For Attributed Networks

Discovering community structures in attributed networks is of great significance and application value for network topology analysis,functional analysis,and behavior prediction.In order to meet the application needs of different fields,researchers have conducted extensive research and proposed a series of community discovery algorithms for attributed networks in recent years.However,the existing algorithms suffer from the problems of insufficient interpretability and ignoring the importance of nodes.To address these problems,this paper conducts research on community discovery algorithms for attributed networks from the perspective of bipartite graph representation and defining node attractiveness.The research in this paper mainly includes the following aspects:In attribute networks,discovering community structures is of great significance and application value for network topology analysis,functional analysis,and behavior prediction.In order to meet the application needs of different fields,researchers have conducted extensive research and proposed a series of community discovery algorithms for attribute networks in recent years.However,the existing algorithms suffer from the problems of insufficient interpretability and ignoring the importance of nodes.To address these problems,this paper conducts research on community discovery algorithms for attribute networks from the perspective of bipartite graph representation and defining node attractiveness.The research in this paper mainly includes the following aspects:(1)Existing community discovery algorithms for attributed networks based on embedding representation map nodes to a joint low-dimensional vector by fusing their topology and attribute information,and then perform community discovery by classical clustering algorithms.However,there is a problem of insufficient interpretability that the relationship between each dimensional vector after the low-dimensional representation is understood.To address this problem,community discovery algorithm for attributed networks based on bipartite graph representation is proposed,which represents the attributed network as a relationship between nodes and representative points.It constructs a bipartite graph using the distance between nodes and representative points,then performs community discovery by spectral clustering.The better range of values for the parameters of the algorithm is tested on synthetic attributed networks,experimental comparative analysis is performed on real attributed networks with existing algorithms to verify the effectiveness of the proposed algorithm.(2)Existing community discovery algorithms for attributed networks based on node similarity first calculate the similarity among all nodes in the attributed network and then perform community discovery by traditional clustering algorithms.However,such algorithms ignore the use of importance information of nodes in the network and don’t consider the use of important nodes to guide the community discovery,so this type of algorithm suffers from high time complexity and high computational cost.Therefore,community discovery algorithm for attributed networks based on node attractiveness is proposed,where node attractiveness is defined by the importance metrics of degree centrality and similarity between nodes.Nodes join the community where the attractive neighbor nodes are located.Experimental comparative analysis with existing algorithms on synthetic attributed networks and real attributed networks is performed to verify the effectiveness of the proposed algorithm.In addition,robustness analysis of the algorithm is analysed.(3)Based on the above two methods,a community detection system is implemented.The system integrates the attributed networks and the research methods in this paper,shows the community division results and provides the visualization of communities.By mining the community structure of attribute networks,this paper helps people better understand the network system in the real world.At the same time,the research results obtained in this paper not only enrich the research content of community discovery in attributed networks,but also provide important technical support for the structure mining in attributed networks.