Research And Application Of Temporal Network Community Evolution

Complex systems in real world can be abstracted as complex networks,and community structure is the most significant structural characteristic of complex networks.Communities often correspond to specific web topics,and to some extent,the functional characteristics of networks are determined by the interactions of the communities.By tracking the community structure,the changes at the mesoscale level can be explored,and the relationship between network structure and the theme can be detected.We can explore the changes of the network at the mesoscale level,and study the relationship between the network structure and the theme.At the same time,the evolution of community also has an important influence on the dynamic evolution of network.Therefore,community tracking is of great importance not only in theoretical studies,but also in practical applications.In recent years,community tracking has attracted significant interests from researchers in many fields,such as computer science,sociology,mathematics,etc.which has become a hot topic in the field of complex network.Based on the analysis of the existing community tracking methods and the shortcomings in these methods,this thesis presents a new method based on activity nodes,which reflects the activity characteristics of nodes in the temporal networks,to analyze the evolution of the DBLP network.The main work of this thesis is as follows:(1)Research the activity of nodes in temporal network.To measure the node activity throughout the network life cycle,this thesis defines the active node window and the node active rate.Then,according to the occurrence of nodes in the adjacent time slices,the nodes are divided into active nodes and inactive nodes.Moreover,we verify the importance of community link and information similarity based on the active nodes.The results show that active nodes play an important role in the smooth evolution of the community.(2)Design and implement a Community Tracking Algorithm Based on active point.This method determines whether there is an evolutionary relationship through the common active nodes to of the two communities in the successive time slices.In the similarity judgment among different communities,we adopt the different similarity functions according to the volume difference between two communities,and the number of common actives nodes to determine the similarity of the two communities.The experiential results based on DBLP dataset show that the algorithm can effectively track the community structure in the temporal network.(3)Study community evolution of DBLP co-authorship.Based on DBLP dataset which is divided into sub datasets annually,this thesis gets the community detection results by the balanced propagation community detecting algorithm according to each year slice and studies the evolution rule.Through the study of the distribution of the six community evolution styles in the data set,this thesis finds that the community has more evolve forms of "appear","disappear" and "expand" in the process of evolution.By continuous tracking of the community,this thesis find that the community life cycle gradually increased over time.Finally,by introducing the text mining technology,this thesis gives the preliminary evolution of network theme,and studies the interactions between the community evolution and the theme evolution.