Research On Local Community Detection Algorithm Based On Line Graph

The concept of community detection has been mentioned in many fields,and it is also an important branch of complex network research,aims to discover the community structure in complex networks.Due to the uncertainty of the community structure on the complex network,the multidimensionality and modularity of the network,and the vagueness of the description of the connection strength,The community has not yet a clear and unified definition in the academic field,but the community structure is objective.The real existence reflects the behavioral characteristics of the individuals in the network within a certain range,as well as the connections between them.Generally speaking,a community is a group composed of nodes whose interconnection density is higher than that of random connections.That is to say,in the same community,the connection probability between nodes is very high,and the edge density is relatively high,while in different communities,the connection between nodes is very few,and even the edges are relatively sparse.Exploring the composition of the community structure and analyzing the relationship within and between communities has important theoretical significance and practical value for understanding and predicting the structure and function of the network.In fact,community detection has already emerged in many practical application field and plays an increasing role,providing important references and clues for discovering,researching and solveing problems.For example: in the biological field,dividing proteins with the same traits into a same community can help us predict the function of unknown proteins and help us predict the function of unknown proteins,analyze the interaction between various protein molecules.After years of exploration and research by scholars,the community structure has gradually expanded from the original non-overlapping community to the current overlapping community.The difference between the two is that in the non-overlapping community structure,a node can only belong to one community,and there is no inter-community.Common nodes,and overlapping community structures can have common nodes,that is,overlapping nodes.Because in a real and objective complex network,only a few communities in the network have no common nodes,so the overlapping community structure is closer to the real world,reflects the characteristics of the real network,and is more widely used.In the past overlapping community detection algorithm,many scholars regard the nodes in the network as the research object,from the node attribute law,network topology and so on,divide all the nodes into their own communities,and get the community partition results of the network.In the recent research of overlapping community detection algorithm,many scholars began to change their thinking,trying to find the community of each side,divide the two vertex of the edge in the same community into one community,and connect the edges of each node There is more than one.When multiple communities of the same node are different,the node is naturally divided into different communities.This method better solves the problem of overlapping communities in the network.The feasibility of dividing the community by using edges as the research object has long been verified and has obvious advantages compared with direct node division,but it is more difficult to divide the edges directly,and the algorithm for dividing the community by nodes in community detection is relatively mature and can be used for reference.Therefore,this paper proposes a local edge community detection algorithm based on line graph.Firstly,we use the line graph model to transform the original graph of the network into its corresponding line graph,and then introduce local community detection The framework combines the degree of neutrality of the nodes and the Page Rank ranking algorithm to divide the communities of the nodes in the line graph,and finally transforms the result of the community division in the line graph into the original graph.In this paper,the normalized mutual information NMI and the extended modularity EQ are used as the community division result index to evaluate and compare other community detection algorithms.The experimental results show that the proposed local community detection algorithm based on line graph has obvious advantages,which verifies this method.Reliability and effectiveness of community division.