Community Evolution Events Detection In Online Social Networks

With the rapid development of Internet,online social networks h ave been integrated into our daily life gradually.The users in online social networks naturally form into group(community)structures bas ed on their social circles and interests.These groups have characteris tics that cannot be observed on single users.Therefore,much attenti on has been drawn to study the group behavior in online social net works.And one of the research focuses is to understand the evolutio n of group structure in social networks.There are two approaches to understanding the evolutional trend of a dynamic network: one is to predict the network development trend by deciding the probability o f individual user behavior;the other is to forecast the network evolu tion by detecting the group evolution events.In this paper,we study the group behavior over time in online social networks,and propose a matrix based community evolution events detection algorithm.Our algorithm includes the definition of dynamic social network events and detection on events.Frist,we propose a “set” and “vecto r” definition of six basic evolution events in dynamic social network s,and use mathematical model to represent these dynamic events.Th en,we propose a concept of correlation matrix which is generated b etween adjacent snapshots of social networks.Based on the correlatio n matrix,decision rules are designed to detect each of the six dyna mic events.Finally,we test the dynamic events detection algorithm b y applying it to two datasets—Facebook dataset and DBLP dataset.The results show that our proposed algorithm can effectively detect t he community evolution events in both datasets.