Event Based Overlapping Community Evolution Detection Method

With the development of information technology.The scale of online social networks is more bigger and bigger.There are virtual communities in many industries,but these communities are changing with time.Such as the community expand,the community shrink,the community split,the community merge and so on.One of the most important issue in complex network research is community evolution.That is,mining the evolution events of community in complex networks with time change.Many reschers have uncovered these potential the evolution events of community from different perspectives.For example,Study the changes in the numbers of nodes in the community in a complex network to measure the types of potential changes in the community,Study the changes of the core structures of community in complex networks to determine what types of changes occur in the complex of community and so on.However,most of the methods are based on non-overlapping community structures and to detect the evolution events of community.But there is a pervasive overlap between community in the real complex networks.However,in the process of evolution of community analysis,the relationships between the nodes in the overlapping domains of the community and the different community of the community are considered.In the process of evolution of community it will lost the information and make the result of the events of evolution inaccurate.At the same time,many researchers have no breakdown of the events of evolution.That is the definition of the events of evolution incomplete.So in this paper,we propose a three way decisions method for the evolution of overlapping communities.In order to find more communities in complex networks,we use three way decisions method for community detection in this paper.This method redefines the concept of community.That is,a communtiy is represented by a pair of set called the lower and upper bounds,Instead of a single set,a community is divided into three regions: positive region,boundary region and negative region.The member in posivive region means it definitely belongs to the community,the member in boundary region means it might belongs to the community,and the member in negative region means it definitely not belongs to the community.In order to analyze some factors that affect the evolution of the community.Firstly we use the community discovery algorithm based on the three way decision to extract the community in complex network according to different time snapshots and analyze the experimental results in detail.Secondly,based on the analysis of the influencing factors,we put forward the definition of community activity and community influence.Finally,we improve the similarity of the community.In order to discover the evolution events that exist in the community over time,we propose an algorthm for overlapping community evolution based on three decision in this paper.Firstly,we according to the similarity of community,community activity and community influence,we determine that there are seven evolutionary events in the community.These seven kinds of evolutionary events are birth,death,expend,shrink,split,merge and stable.Finally,we propose a new algorithm for overlapping community detection according to the definition of seven kinds of evolutionary events.In this paper,a real dateset is used to verify the validity of the algorithm.First,we divide the DBLP dataset according to the time snapshot and based on the three decision community discovery method,the community in different time snapshots is excavated.Secondly,we make use of the community similarity to calculate the matching degree between different snapshots of the community,and further determine the occurrence and evolution of community events in combination with community activity and community influence.Compared with the experimental results in other articles,our algorithm can find out the community evolution events well.It is proved that our method can dig out the community evolution events from another angle.