Research On Node Influence In Social Networks

This paper first introduces the research significance of the node influence in social networks.The existing social networks attract and sink many users with the services they provide.Multi-dimensional information flows rapidly among users.At the same time of high frequency interaction between users,different nodes have different effects on information dissemination,viral marketing,public opinion control and mass incidents due to their different influence.Therefore,it is of great significance to effectively distinguish the influence of different nodes in social networks.From the perspective of information transmission,nodes of different social networks have great differences in the speed,timeliness and scope of information transmission due to the differences in user structure.From the perspective of marketing,it helps new users find relevant information as soon as possible by ranking influential users in different interest directions.From the perspective of public opinion control,the opinions of different users on different platforms have different impacts on the development of public opinion.Fully understanding the influence size and scope of different users has important guiding significance for public opinion control.At present,the specific research is mainly carried out from the following three perspectives.First,it starts from the network structure of social networks,including node information,link information and community information.Second,it comes from the perspective of node influence maximization.Third,specific research should be carried out from the heterogeneous information of nodes.In this paper,different methods are proposed from different perspectives to study the influence of nodes.At first,we adapt the crawler strategy of breadth-first.At the same time,a variety of toolkits were combined to sort out some zhihu.com data,which effectively realizing data cleaning and structured storage.Empirical results verify the basic network characteristics and user behavior characteristics of zhihu.com.Then,combining the idea of node removal and contraction,a centrality method based on node removal was proposed by using the change of the average shortest path of the network before and after node removal.The proposed method was verified by SIR model and Kendall coefficient.Simulations show that the influence of different nodes could be measured more accurately.At present,researchers widely combine different attributes and strategies to mine node influence.Therefore,on the basis of the k-shell method,the location index was processed with sigmod function to measure the location attribute of the node.The local information of the node was used to measure its neighbor attribute.The two kinds of attributes were weighted by information entropy.Simulations verify that the improved k-shell method based on multiple attributes can significantly improve the performance of identifying the influence of nodes.Finally,based on the idea of heuristic algorithm and the similarity index of node pair,we proposed a node influence discovery algorithm based on community partition.The algorithm takes the center nodes of multiple communities as members of the influence node set.Experiments show that this method has better performance than partial heuristic algorithm.To sum up,centrality method,multi-attribute method and heuristic method all measure the influence of social network nodes from different perspectives.The three methods proposed in this paper have certain effects on the mining of influence of nodes.With the rapid development of information technology,the information dimension of single node increases,the interaction between users becomes more complex,and the network heterogeneity becomes more diversified.All these presents new and huge challenges for the research on the node influence in social networks.