Research On Role Recognition Of Complex Network Nodes Based On Structure And Property

With the great progress in theory and practice,the application of complex networks has been involved in all aspects of the society.As one of the important issues in the study of complex networks,the research on the role of nodes is gradually concerned by people.According to the research on the role of nodes,it is not only the study of the nature of a single node,but also the relationship between multiple nodes,and so on.The recognition of the role of nodes becomes an important aspect in the study of complex networks.It has great practical significance in the field of the network dynamic analysis,anomaly detection,image processing,biomedical research,crime organization disruption,public opinion leader mining,rumor source detection.This paper focuses on the role identification based on node structure and attribute,and puts forward different methods of character recognition and role analysis according to different emphases.(1)A static and dynamic character recognition method based on directed weighted topological potential :The concept of topological potential was applied to the weighted network.Based on edge weights and node attributes,a measure of the node role identification was constructed.In this paper,the influence factors were discussed,and the typical role was defined according to the discrepancy of the nodes.Then the method of static node role identification based on distance division was proposed.Considering the change of the role of the nodes in a long time,the role of nodes in the sequence was analyzed from two aspects: the role function of the node and the pattern of role-change.(2)Analysis of network evolution from the perspective of the change of the roles: Taking the dynamic nature of the network as the premise,from the point of view that changeof node roles reflects the perspective of network evolution,the network evolution process was analyzed in detail.We analyzed the evolution of the network and the important change points in the evolution process by four different aspects,which are the distribution of the topology potential of nodes,the role distribution ratio,the network topology potential entropy and the node's dependent probability to different roles.(3)This paper also focused on the role of the online social network node,which according to the vast amount of information in the online social network to dig the role of the users.First,we made an empirical analysis of the user characteristics of online social network.Taking the user data of micro-blog network as the object,we did the work on statistical analysis,correlation analysis and empirical analysisof the network user's basic attribute and user behavior,which was to provide the basis for the construction of the next user feature vector.(4)in this paper,a method of online social network user role identification based on SVD-NMF was proposed.According to the result of the characteristic quantity analysis and many concepts of sociology,we constructed the user characteristic vector and the characteristic matrix.The initialization of NMF was optimized based on SVD algorithm,which determined the rank and initial matrix.At last,we proposed the method of division of users based on this SVD-NMF algorithm.