Research On Community Division And Maximization Of The Community Based On Node Similarity And Distance

Social network is an important information dissemination platform in today’s era,and a major channel for people to express their opinions,share their feelings and interact and communicate.In social networks,some users have high influence and can attract more followers,trigger more discussions,and even influence public opinion and social change.Therefore,how to find the influence of a certain user,organization,brand,event,etc.in social networks has become an important and challenging problem.The main research of this paper is as follows:(1)Node similarity and path distance based community discovery algorithm(NSPD algorithm).In order to more fully consider the influence of topic networks on community partitioning,the algorithm uses topological structure and topic semantic information to calculate the attribute similarity between nodes to construct a social network model,and improves the K-means algorithm by combining node attribute similarity with path distance on this basis to improve the accuracy of overlapping community discovery.In order to verify the performance of the algorithm,this paper conducts comparison experiments with four comparative algorithms on four datasets,and shows that the algorithm can accurately and effectively perform community segmentation and guarantee the balance of running time and segmentation results.(2)Community discovery based influence maximization algorithm(CDBIM algorithm).The algorithm,in order to better fit the sparse structure of the real network,performs intra-community decomposition in the delineated thematic community structure,combines similarity and path distance to calculate the kpi value to improve the kernel algorithm,so as to find the core nodes and select the nodes with higher similar average influence into the candidate seed set.Then this paper adds node similarity to the calculation of propagation probability,improves the IC propagation model to make the whole propagation process more realistic,and uses the CELF algorithm to obtain the final seed set on this basis.This paper verifies the reliability of the CDBIM algorithm by conducting experimental comparisons with three comparison algorithms in six different datasets.In summary,this paper proposes an overlapping community discovery algorithm based on the full consideration of social network topic information,which has a better community division capability by considering both node similarity and path distance.And based on the community division results,we propose an influence maximization algorithm that balances the operational efficiency and influence result size by seed node selection within each community.