Research On Analysis Method For The Evolution Process Of Communities In Dynamic Networks

A large number of networks in the real world are constantly changing over time,so they are also called dynamic networks.In order to explore the huge value hidden in these dynamic network data,dynamic network evolution analysis has become a current research hot spot.Among them,community evolution analysis in dynamic networks is one of the most meaningful research directions,which has important values in applications such as influence analysis,information dissemination,and network marketing.On the basis of summarizing the existing research on the analysis of the evolution process of the dynamic network community,this paper conducts in-depth research on the important stage in the basic framework of the analysis of the community evolution process.The main work includes:First,in the dynamic network time slice dividing stage,aiming at the inaccurate community evolution process caused by fixed-length time slice partition,a time slice partition method based on edge set similarity was proposed.This method dynamically adjusts the length of time slice by calculating the influence of update operation on the similarity of network edge set.Experimental results show that this method can reduce the problem that the fixed-length time slice can’t effectively extract the characteristics of network structure when the network structure changes dramatically,which leads to large differences in community structure detection results,and covers more community evolution events in the evolution process.Second,in the event matching stage of community evolution,in view of the existing community evolution process analysis methods,some important evolution information will be missed when describing the community evolution process,so a community evolution process characterization method based on cross-time slice is proposed.According to the relationship of evolutionary events between communities,this method dynamically increases the matching range of time slices by crossing time slices.Experimental results show that this method uses more community events to describe the process of community evolution,and can find some hidden community evolution relations,which provides richer semantic information for studying the law of community evolution.Third,combined with the previous research content,a community evolution process analysis system was designed and implemented.The system can perform community structure detection on the input dynamic network data,and then perform evolution analysis on the detected community structure,and visualize the results.It is convenient for users to analyze the evolution process of community structure.