Research On Community Discovery Algorithm In Complex Networks

Community discovery has always been a hot research topic in complex networks and their applications.With the rapid advancement of complex network research over the years,research on community discovery algorithms has attracted widespread attention from scholars in various fields.Its related research has been widely used in physics,sociology,biology,medicine,psychology,broadcast media and other disciplines.On the one hand,through the accurate division of network structures,which let us to better understand its internal structureand attributes by learning community discovery algorithms,and it also has a deeper understanding of the various network structures that exist in life.On the other hand,improving the accuracy of community discovery,enhancing the robustness of the network structure,optimizing the network structure,or has a particularly critical theoretical value and application significance for the commercial division of online consumer groups.In many community discovery algorithms,label propagation algorithm stand out with numerous advantages.However,in the process of implementation,there are certain flaws.The main idea of this topic is to address the shortcomings of low accuracy and poor stability in the random selection of the label propagation algorithm.After improving the label propagation algorithm,two algorithms are proposed to give full play to the centrality of the node.The role of community discovery is to reduce the error of the results of algorithm partitioning caused by random selection and updating of label by ignoring the influence of the weight of the node itself.The main idea of this paper is to address the defects of randomness in various stages of the traditional tag propagation algorithm.The corresponding improvement methods are proposed for the implementation process of the algorithm,and the role of node centrality in community discovery is fully exerted.Measure the size of the node's influence in the network range,and reduce the random selection probability that the algorithm is treated equally in the implementation process because it ignores its own attributes.The main tasks include:1.The concept of local important nodes and node centrality is integrated into the idea of algorithm improvement,labels are assigned to selected local important nodes,and then label updates are performed.From the evaluation indicators of the experiment,we know that the improved algorithm is slightly better in terms of accuracy and stability than the results of several related algorithms.2.The concept of Leader Rank centrality is introduced into the improvement of the original LPA algorithm.The important nodes filtered according to the characteristics of the Leader Rank method are arranged in descending order.Locally important nodes are locked into the propagation source,and other strategies are used to start the propagation of the labels.According to the comparison of the experimental results,the improved algorithm and the comparison algorithm presented in this paper highlight certain advantages in the relevant evaluation indicators,and the accuracy rate is greatly improved.