The Empirical Analysis And Application Of The Information Propagation In Online Social Network

Micro-blog,one of the classical online social applications,has become the main platform to get and spread information in the modern information age.Therefore,to study the structural characteristics and information spreading mechanism of micro-blog will help us to understand the evolution of network and monitoring the information spread.In this paper,Sina Micro-blog is taken as the main research platform and the data of users and information propagation is dug and analyzed.Furthermore,the influence of users' behavior and network structural characteristics on information propagation is discussed.The main contributions of this thesis are as follow:(1)To obtain the Sina micro-blog' users data and different kinds of micro-blog messages,a web crawler is developed.Based on a completed message spread process,users' relation network and message propagation network are built.Through the empirical analysis of two networks' structural characteristics,it is found that both networks has similar features,such as small world property,scale-free distribution on nodes' degree and low reciprocity.In the analysis of assortativity coefficient,the results show that users' relation network is assortative,but the latter has no assortative feature which implies that information spread has influence on the evolution of message propagation network.(2)Four kinds of micro-blog messages are taken as research object and a rumor spreading model is proposed.Firstly,different kinds of messages' propagation features and the influence of users' behavior on the messages spread are discussed.The results shows that the temporal distribution of retweets on micro-blog has the same characteristics of long tail.It is found that the users' comments when they retweet has influence on rumors propagation and causes fluctuation of temporal distribution curve.Besides,the behavior of secondary retweet occurs frequently which increases the probability of reading by followers.Based on the empirical analysis above,a new rumor spreading SEIAR model is proposed with considering the behavior of retweet and comment.This model is utilized in the simulation on a real network and results show that SEIAR model conforms to propagation characteristics of real rumors in micro-blog.(3)A new algorithm in identifying vital nodes is proposed by using HyperMap method and it extends the application of hyperbolic space theory in the field of complex networks.According to the existing results of hyperbolic space,this thesis applies growing network model E-PSO and HyperMap method to map the real network to hyperbolic space and verifies the performance of HyperMap method in link prediction area.Then,based on the definition of hyperbolic distance of two nodes,a new vital nodes identification algorithm HCC is proposed by modifying closeness centrality algorithm.It is found that HCC algorithm performs well in high propagation probability condition when compares to other algorithms.