Research On Social Network Modeling And Mining Algorithms

Social Network Analysis(SNA)is a research branch which has been gradually developed in the fields of sociology,psychology,anthropology,mathematics and communication science since the 1970s.Social network analysis is not only a tool,but also a way of thinking about relationalism.For example,some researchers in the field of information science have applied social network analysis methods to develop a series of researches on competitive intelligence,knowledge management,library resource allocation,hot topic analysis,citation analysis,collaborative research analysis and blog networks.In essence,social network analysis belongs to the field of complex network theory,and it is a typical interdisciplinary.The application of social network analysis can not only rationalize the distribution of resources,but also improve the efficiency of information dissemination,and better achieve the optimal allocation of labor market resources.Ultimately,it can guide people to understand the nature of social phenomena,to guide people to predict the development of social networks,so as to solve the actual social problems.In this dissertation,we adopt the social network of the workplace as the object tocarry out research works on social network modeling,topology characteristics analysis and evolutionary characteristics analysis and invulnerability analysis.Aiming at the existing social networks and the social network models proposed in this dissertation,the node importance analysis and key link mining algorithms are studied.Concretely,this dissertation starts with the study of various social network models,and constructs the network structure with certain characteristics through simple and refined models.These models transform complex social network problems into simple mathematical problems.Then,the topology characteristics and evolution characteristics of the network models are studied.Nest,the dissertation studies the node importance and survivability of the social networks by using various centralities,and proposes a key link mining algorithm.The main innovative works done in this dissertation are as follows:First of all,this dissertation presents a social network model that reflects the change of the workplace environment.The social model of the workplace can be used to study the social relations of people in the workplace more simply.In this model,we consider the node as an individual,the connection as a social relationship,and the group as a social organization.The model uses two parameters,one for the exchange rate that stands for the strength of the connection in the organization;the other for the quit rate,that denotes the nodes' jumping frequency between organizations.Thus,we simplify the complex social relations to the mathematical model,and according to the mathematical model,we study some topological characteristics of the social network of the workplace.The simulation results show that the network generated by the workplace model presented in this dissertation has the small-world characteristics.By setting different parameters,the resulting networks have different clustering coefficients,group distributions and degree distributions.Secondly,in view of some of the mechanisms used in the above-mentioned social network model being not very reasonable,by drawing on some common social phenomena,we put forward an improved social network model that is more close to the reality of the workplace.Based on the exchange rate and the jumping rate,we adopt the community structure to build a social network model for the first time.The simulation results show that the improved model has better realism,larger mean degree distribution,smaller mean distance and larger clustering coefficient than the previous model because of the higher mobility of nodes.It is shown that the improved model is in line with the social characteristics,which is helpful to study the social network.Then,this dissertation makes use of various centralities to study the importance and invulnerability of nodes in social networks.In this dissertation,we focus on the performance comparison in guiding the social network attack between the traditional centralities(the degree centrality,the betweenness centrality and the closeness centrality)and the centralities proposed in the literature in recent years(the second-order centrality,the Laplacian centrality and the total information centrality).The results show that,Finally,this dissertation presents a key link mining method based on entropy weighting method and gray relational analysis for social network.In order to evaluate the importance of links in complex networks,it is critical to select the evaluation criteria in a comprehensive,objective and independent manner.The method proposed in this paper utilizes the advantages of gray relational analysis and entropy weight analysis in multi-index fusion analysis.The comparison between the two traditional social networks shows that the method is more applicable and more comprehensive than the other traditional methods.