Study On Identification Method Of Key Role In Mass Incident Based On Social Network

Our country is in a period of important strategic opportunities for social development.The rapid development of social economy is accompanied by prominent social contradictions and frequent social mass incidents,which bring many new problems to social public safety and crisis management and new challenges to the construction of a harmonious society.With the popularity of the Internet,information on various social platforms shows the characteristics of fission and multi-center dissemination.Social platforms have become the main platforms for public opinion dissemination of mass events.How to effectively control social platforms has become the key problem to be solved urgently in the construction of a harmonious society.Taking social network data as the research object and identifying key roles as the research objective,in this thesis,the relevant theories and methods of identifying key roles in mass incidents are studied.The main works of this thesis include the following aspects:(1)Based on Forces-Oriented Layout model,a new community detection algorithm is designed in this thesis.First,on the basis of Forces-Oriented Layout model,the definition of attraction and repulsion of nodes is improved.Then,combined with clustering algorithm,the algorithm proposed in this thesis solves the problems that the community structure is not obvious in the force oriented algorithms and the visualization is not satisfactory in traditional community detection methods.Finally,experiments on multiple public data sets verify the superiority of the algorithm in terms of indicators such as NMI and modularity,and successfully identify the community structure in mass incident,providing a data basis for identifying local key roles in social networks(2)According to the theory of entropy weight theory,a new method to calculate the influence of nodes in social network and an algorithm to identify key roles are designed.The method of influence calculation integrates four importance evaluation indicators of nodes: degree centrality,betweenness centrality,proximity centrality and clustering coefficient,effectively avoiding the limitation of using a single indicator.Then,according to the node's global and local influence value,the nodes are clustered by K-means clustering algorithm,and key roles in the network will be identified.Comparative experiments with multiple public data sets verify the feasibility of the algorithm in this thesis and the key roles in the mass incident data set are identified.(3)A new calculation method of key role evolution is proposed after calculating the influence of nodes in the sequential networks.On the basis of community detection and role identification,the changes of key roles are mined by continuously comparing the role characteristics of nodes in each time slice.The results of experiments show that many key roles whose influence gradually declines over time no longer have effective influence on other nodes.There are also some nodes whose influence gradually increases become new key roles with the development of the situation.