Research On Dynamic Network Community Detection Modelling And Evolution Feature Analysis

Community-scale fraud,advertising and other behaviors are representative of new important security risks in the urban system.In the context of computational sociology,based on mobile phone signaling data,using the structure mining and dynamic modeling theory of complex networks,combined with urban community characteristics and background data,can effectively realize the detection,evaluation and prediction of such new risks.Due to its powerful representation and prediction capabilities,dynamic complex network community detection and evolution analysis is one of the core contents of network science research and application.The current dynamic network analysis focuses on how to improve the detection results of the community,ignoring the association between the dynamic evolution behavior of the communities,nodes and the intrinsic characteristics of the network structure.This paper designs a study of the core driving factors of community evolution in dynamic networks based on the statistical structure characteristics of nodes and communities.And then we use the statistical network model to model the multi-scale network evolution behavior.Furthermore,we use the evolution model to make a case study to utilize the node evolution attribute into urban risk analysis and prediction by using the phone call data.The main work and research contributions of this paper are as follows:Firstly,the structural properties of nodes and the laws of community evolution are explored.Through clear design,we regard whether the nodes in the community are transferred at the next snapshot as a binary-classification problem,using the feature engineering to extract the structural attributes of each node as the classification feature,and using the decision tree for the binary classification of social media data,paper reference data and so on.And the community transfer of nodes is classified and analyzed to figure out nodes characteristic importance in its community transfer.Secondly,the community detection model HB-DSBM,which combines the DSBM with the evolution characteristics of the nodes.HB-DSBM combine the node-level community transfer trend and the community-level transfer trend,constructing the hierarchical Bayesian dynamic stochastic block model by constructing the hierarchical Bayesian structure from the community-level transfer matrix to the node-level community transfer matrix.Then we estimate the parameters of the model using variational inference.Finally,the mobile phone signaling data is used to empirically analyze the above rules and the validity of the model in urban risk calculation.The mobile phone call data is used to verify the degree of the node and the average neighbor degree of the node to analyze whether the node community belongs to the transfer and find a reality.At the same time,the effectiveness analysis of the hierarchical Bayesian dynamic random block model is carried out.The processing of the real network is used to extract the node community membership and community evolution information in the real network,and then the real-world events are extracted and targeted by the community detection result.The potential risk of the data is analyzed and visualized.This paper expands the new ideas of community detection and modeling in dynamic complex network research,expands the application of statistical models in network analysis,and provides a theoretical paradigm and new direction for urban risk calculation.