Research On Edge Community Discovery Algorithm Integrating Integrated Learning Strategies And Line Graph Models

With the rapid development of internet technology and social networks,exchanging information and sharing opinions online has become the norm.One critical challenge is how to extract valuable knowledge from the vast amount of data generated.Community structure,as one of the most fundamental features of complex networks,can help us understand the topological characteristics of the entire network and the relationships among nodes.For example,nodes with similar topological structures have higher similarity and are more likely to belong to the same community.Continuously exploring community structures in networks can help us better understand the internal rules of communities and the topological structure of the network.In recent years,the issue of community discovery has attracted widespread attention from scholars at home and abroad,yielding fruitful results.However,the field still faces key challenges,such as the difficulty of dividing nodes with overlapping attributes in community discovery and using the topological structure of edges within communities for community discovery.Line graphs are a widely applicable model for edge community discovery,which can naturally use the topological structure of edges for community discovery.Based on these issues and theories,this paper delves into and improves upon the shortcomings of some existing classic community discovery methods,and proposes two new community partition models.The specific work is as follows:(1)This paper proposes a stacked ensemble learning strategy based on a line graph ensemble learning overlapping community discovery algorithm(OCD＿SEL),which combines a community detection method on an online graph with a multi node clustering method.Our method is mainly divided into the following parts: firstly,a new local community discovery algorithm(LCDDCE)is adopted on the online graph based on degree centrality,which determines the final number of communities.Then,several node clustering methods are combined to classify the nodes,and multiple communities are obtained by transforming the community into a community on the line graph and locating the local center point of the community on the line graph for local expansion.Then,the number of basic clusters is determined based on the generated number of communities,and the line graph community is transformed into overlapping communities.Finally,the final node category is determined by voting on various clustering methods.(2)An improved algorithm for discovering overlapping communities in edge communities based on the Matthew effect is proposed in this paper.Inspired by the Matthew effect,an algorithm called(OCDME)is proposed,which combines the Matthew effect in human society and treats the line graph as a social system for discovering overlapping communities.This algorithm uses the similarity matrix on the line graph to replace the original node’s degree matrix,and filters overlapping nodes from the perspective that they should contain neighbors from different communities,thereby improving the accuracy of community partitioning results.Finally,the two algorithms proposed in this paper are experimented on three real-world datasets and one biological dataset,and the experimental results obtained from the algorithms are compared and analyzed with the experimental results of existing overlapping community discovery algorithms to verify the effectiveness of the algorithms proposed in this paper.