Generation Model For Dynamic Complex Network Based On Node Popularity

With the rapid development of modern technology,more frequent human interaction hides more risks,so it is significant to detect risk groups and individuals.The mobile phone data with high precision and time information is used to build a dynamic network,and dynamic community detection method is used to identify community structure,evolution tracking,node importance and node role,to realize the functions of risk group identification and risk personnel identification in urban risk management.The existing works for dynamic networks analysis which are based on the stochastic block model always analyze the evolution of dynamic networks by introducing a probability transition matrix.The nodes within the community are assumed as undifferentiated,and the heterogeneity of nodes and the scale-free characteristics of the network are ignored,so these methods cannot accurately model the real-world network.To overcome these limitations,I based on the stochastic block model,study the generation model for dynamic complex networks with node popularity.The work is as follows:Firstly,a dynamic network generation model with scale-free characteristic(DPSBM)is proposed.Aiming at the property that the degree of nodes in real networks obeys power-law distribution,the scale-free attribute of real networks is simulated by integrating the heterogeneity of nodes,the deviation of existing models in community detection is corrected,and the dynamic network in real-world is modeled more accurately.At the same time,an effective variational EM algorithm is proposed for model inference.Secondly,a dynamic network generation model(DRSBM)combined with node topological features is proposed.The idea of using single popularity to model the difference of nodes is improved,and the dynamic complex network is constructed by integrating the topological features of nodes,so as to model the internal relations of the community assignment,roles of nodes and topological characteristics,and distinguish the different functions of nodes in the network.Community detection,evolution tracking,and node roles are solved under a unified framework,and an effective algorithm based on mean-field theory and variational EM is proposed.Lastly,the case study is based on mobile phone data.Aiming at the problem of fraud identification in urban risk,we use the dynamic community detection method proposed in this paper to detect community structure,track evolution,and analysis the importance and roles of the individual.At the same time,identify the individuals and groups with criminal risk,and providing strong support for urban risk management.In Summary,the generation model of dynamic complex network proposed in this paper integrates the node popularity,corrects the deviation of the existing model in community detection and evolution tracking,and finds the role and importance of the nodes in the network,which is more suitable for the real-world network,and expands the application of the stochastic block model in the real network.