Research On Community Detection Algorithms In Complex Networks

Complex networks are powerful tools for describing many systems in the real world.Complex networks can be represented by graphs.In the representation of the graph,the nodes represent independent individuals in the network,and the edges represent the connections between individuals within the network.Complex networks have a distinct community structure,which is characterized by tight connections between nodes within the community and sparse connections between communities.Community detection provides an important way to understand the structure and function of complex networks.Discovering communities in complex networks is of great significance and broad application not only for analyzing the topological features of complex networks,understanding the functions of various parts of complex networks,discovering potential laws in complex networks,and predicting the behavior of complex networks.prospect.The topology of a complex network contains a lot of information.However,many traditional community detection algorithms use only part of the network structure information when performing community detection,and there is a waste of information.In order to make full use of these structural information,this paper integrates more network topology information into the process of community detection from the perspective of traditional community detection and network representation learning,thus improving the community detection effect.With the rise of community detection,a large number of scholars have invested in the field of community detection,and proposed a variety of community detection algorithms,which greatly promoted the rapid development of the field.Among them,the label propagation algorithm has received extensive attention due to its ability to discover community structures from large-scale social networks.However,the stochastic strategy adopted by this algorithm will not only reduce the efficiency of the algorithm,but also bring instability to the results of community partitioning.In order to solve the problem caused by random strategy,this paper proposes a label propagation algorithm based on local optimization.The algorithm introduces a pre-propagation mechanism to prepropagate tags according to specific factors(such as node tightness)to achieve the purpose of optimizing randomly initialized tags.After that,the node tags are traversed and updated according to the descending order of the aggregate influence.The experimental results fully demonstrate the effectiveness and usefulness of the proposed algorithm.In recent years,machine learning methods have made breakthroughs in many fields,and some scholars have tried to apply the methods to community detection.However,each machine learning algorithm has its own advantages and disadvantages.Which algorithm is more suitable for community detection is still an open question.This paper will use the machine learning algorithm to learn the feature representation of the nodes and discover the community structure by clustering the node feature representations.By referring to the classic model Word2 vec in the field of natural language processing,this paper introduces a topologically sensitive negative-class node sampling method in the negative sampling process,so that the node representation can better reflect the global structure information of the network.The node representation with stronger expressive ability is obtained,and then the nodes are clustered according to the CFDP clustering algorithm to discover the community.The experimental results show the effectiveness of the method.