Research On Link Community Identification Method Based On Line Graph

In recent years,with the deep study of networks,the network overlapping structure has attracted more and more attention from scholars.At the same time,identifying overlapping structures in the network has also become a hot research topic in the field of data mining.So,as an essential means to identify overlapping structures,overlapping community detection becomes the leading way to analyze the relationship between communities and contribute to the study of complex networks.On the one hand,it plays a vital role in the process of mining correlation between commodities.On the other hand,it can effectively reveal the hidden overlapping structure in the network.Among them,the research method based on the line graph is one of the commonly used research methods in overlapping community detection.Therefore,the paper proposes the following two overlapping community detection methods.1.Link community detection based on ensemble learning(LCDEL).LCDEL algorithm first proposes a novel distance metric,which measures the distance relationship between nodes by calculating the covariance between nodes.After that,the algorithm gets link communities according to the combined k-means clustering algorithm with ensemble learning.LCDEL algorithm reduces the adverse effects of k-means random initialization of clustering centers and achieves the goal of optimizing the algorithm's results.The algorithm has been tested on several real networks,and the experimental results verify the effectiveness of the algorithm proposed in this paper.2.Link community detection combined with network pruning and local community expansion(NPLCE).Based on the relationship between the network topology and redundant nodes,NPLCE algorithm proposes a link attractiveness formula to delete redundant links in the network.In addition,when calculating similarity on the line graph,common neighbors between connected nodes and their connection tightness are not entirely related.So,NPLCE algorithm proposes a novel local community discovery framework based on the topological relationship between non-adjacent nodes to find local communities in the line graph.NPLCE algorithm has been tested on multiple real networks,and its accuracy and effectiveness have been verified.There are two methods proposed in this paper.The first method is research-based on ensemble learning.It mainly solves the problem that k-means clustering affects the accuracy of algorithm results.The other is a research method combined with the topological relationship on the line graph.The method mainly solves too many redundant nodes in complex networks and the problem that adjacent nodes with multiple common neighbors are not close.Both methods are well verified in multiple real networks.