The Research Of Community Discovery Methods Based On Node Influence

In the study of social network-related issues,community discovery has always been an important role.In recent years,the development of computer network technology makes the scale of social networks extend from the relationship of people in real life to the online virtual network.As social networks themselves become larger in size,the overlapping community structures they present become more complex.In the existing community-discovered algorithms,the importance of node influence in the community discovery process is neglected to a certain extent.In this paper,a method KSADWN for calculating node influence is proposed,which combines the global influence,local influence and the contribution of the neighbors comprehensively,based on the K-Shell decomposition algorithm for the shortcomings of the coarse grained partitioning and lack of consideration of other factors in addition to the global situation.The experiment is validated on five datasets of different sizes,and seven classic influence algorithms are used to prove that the KSADWN algorithm proposed in this paper has a good performance.Based on the calculated node influence,a overlapping community discovery method INILPA related with the of label propagation algorithm is proposed,which is to improved the shortcomings of the randomness and lack of consideration node influence in the computation for the COPRA algorithm uesd to discovery the overlapping communities process.The label update sequence is improved during tag propagation,and not only the influence of the neighbors are considerated but also the attributional tendencies include own label factors in the label spread process,the own influence of the node is also considerated.The INILPA algorithm improved the method of calculating the belonging coefficient of the label,and fusing the influence of individuals to improve the setting of the overlapping threshold thresholds.The experiment is validated on five datasets of different sizes and uses six classic overlapping community discovery algorithms for comparison.It is proved that the proposed INILPA algorithm can effectively reduce the randomness of the algorithm and improve the stability of the algorithm.,it can improve the quality of community discovery.