Structure Analysis Of Large Graph Data Based On Minimum Description Length

Cyberspace data increases with each passing day in the era of rapid development of the Internet of nowadays.These data of large amount include the privacy and informa-tion of people in real life.Extracting the information of these data through data mining method of cyberspace and mining the privacy information that the data included is an important part of current social network structure analysis and cyberspace security re-search.In demand of theoretical research and application,large-scale network data need to be compressed.Process topological structure mining.Analysis and understanding its topological semantics.This thesis studies the structural data and the problem of structural analysis of graph topological data,in cyberspace under the current security background.Emphasis is laid on overlapping structure mining,semantic recognition and analysis of subgraph structure and the structure mining of graphs.The graph analysis method in this thesis is mainly applicable to social networks,but it can also be applied to other network types with power-law distribution.The minimum description length principle is introduced as a criterion for subgraph clustering,subgraph semantic structure recognition and graph structure model evaluation.In order to mine the multi-structure uniformly in graph,this thesis mines several kinds of subgraphs through their short diameter commonness,obtains the decomposition of graph and the set of subgraphs by the subgraph decomposi-tion algorithm based on hub-node and the sub-structure aggregation algorithm based on egonet.Then,the structural semantics of subgraphs are identified and mined uniformly through the minimum description length principle and graph encoding algorithm.Fi-nally,with the minimum description length,the topological structure summary of the graph is obtained by substructure combination.Through the research of this thesis,the structure of the graph is mined and analyzed,and the structure of the network is un-derstood.Experiments show that the proposed method is a good compression,mining and understanding of the overall topological structure of graphs compared to current method.This thesis makes full use of the minimum description length principle.The condi-tions of graph merging and overlapping relationships between subgraphs are determined by the minimum description length principle while mining subgraph structures.In graph encoding and semantic structure mining,the minimum description length is still used to help identify the type of subgraph,and a tree structure template for sparse structure is introduced to improve the current graph structure summary algorithm Vog.