Research On The Temporal Social Network Community Detecting Algorithm

In recent years,the study on social network has attracted more and more attention.It has great significance and far-reaching impact in computer science,physics,biology,and management science.Scientists have a certain understanding on the static characteristics and network topology of social networks,along with the deep of the study,scientists have begun to do researches on dynamic social network with time,and studies on micro-community evolution characteristics are becoming more and more concerned.Community detection is the essence of data mining clustering algorithm applied in the social network,community detection has important significance on topological structure and functional characteristic?information recommendation and network behavior prediction Based on the important significance of temporal social network community detecting,in this thesis,the main research content includes three parts: temporal social network,community detection algorithm and community evolution.Firstly,this thesis briefly introduces research background and the significance of the temporal social network and community detecting.And then it analyses basic concepts of social network,social network modeling and properties,completing sub graph generation algorithm technology.This thesis uses community detectingalgorithm CPM,which based on complete subgraph,analyzs the core idea of CPM,uses real data sets and simulation data sets to do the experiment and analyses results of CPM,and gives the realization of the serial and parallel implementation,analyzes two kinds of time efficiency realization,and put forward the method to improve the shortcomings of the disadvantages of CPM.It uses scholarscooperation network DBLP dataset to cut connected subgraph and puts forward connected branch pedigree tree construction algorithm and uses the main connected components to construct temporal social network,and uses the core architecture for CPM community detectingalgorithm to apply in the temporal social networks to find community.Using twelve sets of experiments to find out two optimum parameters could make the effect of experiment the best,and using prior knowledge to validate the results.Meanwhile,it studies DBLP community evolution.Finally,it proposed the algorithm of acquisition community evolution result and analysis community evolution branch results from the four aspects.Inthis thesis,the experiment results show that: the parallel CPM realization has highertime efficiency compared with the serial mode,especially in the case of a large amount of data.Taking DBLP as the experiment data of community evolution,twostates including single morphological evolution and integrated community size average growth rate under integrated and continued evolution is higher than that of single morphological evolution,not BA community in the process of community evolution becoming BA community.This thesis enriches the research results in this field to a certain extent.