The Study Of Self-excited Hawkes Process Based Group-popularity Prediction In EBSN

In recent years,traditional online social networks which have occupied an important place in our daily lives are no longer able to meet people's further social needs.In such case,as a new type of heterogeneous social network with special features,EBSN(Event-Based Social Network)has come into being,which not only provides an online platform for idea exchanging and experience sharing,but also helps users to participate in offline social events in group.With the continuous development of EBSN,issues related to EBSN have also been widely research by scholars.Apart from other issues like content recommendation in EBSN,content popularity analysis and prediction also attracts people's attention.Therefore,this paper studies the popularity prediction problem of the group which occupies an important position in EBSN,and then analyzes the issue and proposes a prediction algorithm to solve this issue.Firstly,the EBSN network model is constructed based on the data analysis,followed by relevant definitions of group-popularity of EBSN.Then on the basis of the above,this paper analyzes and extracts corresponding features from four aspects:inherent characteristics of the group,historical trend of the group,internal emotional feature and external dynamic feature.And then this paper constructs the group-popularity model based on the Self-Excited Hawkes Process.Furthermore,an EBSN-Hawkes Process Algorithm(EHP)is proposed based on the idea of stochastic process in EBSN.This algorithm not only takes into account the influence of group's inherent characteristics and its historical trend on the prediction of group-popularity,but also integrates the contribution of internal emotional factors and external dynamic factors to improve the effect of group-popularity prediction.Finally,experiments are carried out based on a real EBSN dataset.Through the analysis and comparison of the experimental results,we can conclude that the EHP algorithm proposed in this paper can effectively predict the group-popularity in EBSN.Besides,compared with other contrast algorithms,EHP algorithm has better effectiveness.At the same time,external dynamic characteristics of EHP algorithm contribute more to EBSN group prediction than inner emotion feature.