A Directed Tie Strength Estimation Method Based On User Characteristic Attribute,Topology Of Networks And Social Interactions

As the product of web2.0 era,the emergence of social network sites make people bridge the traditional social gap caused by the temporal and spatial distance.Its wide spread and rapid development have changed our lifestyles dramatically.With the mass creation and frequent exchange of User Generated Content(UGC),an amount of interactive data and complex user relationship have been generated in the social network sites,this phenomenon attracts the attention of enormous researchers.The information posted by user,like pictures,texts,videos,is a potential and rich resource,which reveals the real selves of users.Therefore,how to dig the hidden economic value has become scholars' research hotspot.However,extant researches about the tie strength mainly focus on the user attributes and social interactions.They ignore the influence of network structure,and don't take into account the direction and the reciprocity of social interactions.In addition,previous studies mostly pay attention to the binary relationship of users,and mainly adopt qualitative research methods,so the researches about the relationship strength between users are too shallow.Hence,this thesis proposes a dissymmetrical tie strength estimation method based on user attributes,topology of networks and social interactions(DSTS-ATI).The main innovation points are as follows:(1)Social network user attributes mainly has two types:background profile and dynamic online status characteristic.Background profile is individual's social identity,such as education background,which reflects users' social and economic status.The dynamic online status characteristic is an accumulative state,which is produced by the complex relationship and the social behavior.Besides,user's text message can partly reflect the user's interests and preferences.Thus,when calculating the similarity of the user attributes,the author considers the static characteristics,dynamic characteristics and text message.(2)In the process of analyzing the online social network topological structure,this article considers the direct and indirect relationship between the users,and analysis community structure based on the nodes and paths between users.More detailed,this thesis utilizes common neighbor nodes and the links between common neighbor nodes to figure up the direct relationship between users,and consider the paths between nodes and the weights of paths to measure the indirect relationship.(3)The difference between the follow-ship between the users,results in the difference of initiative and passive of the social interactions.This phenomenon is bound to affect the strength of the relationship between the users.Therefore,the behavior of considering the only one side of the relationship to measure the intensity of relationship,or only focusing on the user who are each other friends,is not considerate.To this end,this thesis gives out a directed interaction strength calculation method,which pays attention to the perception of relationship on both sides.(4)Besides,in order to comprehensively measure the effect of the strength of social interaction,while considering the frequency of social reciprocity behaviors,this thesis take into account the different importance of social behaviors at the same time.What's more,the strength of users' interaction is the dynamic,with a short-term smooth phenomenon.Based on this,when calculating the strength of interaction,this thesis introduces time slice and the delay factor into model.This thesis takes microblog users as research objects,and gets access to user data through the crawler.In order to make sure whether the method is scientific and reasonable,the article designs a series of comparative experiments,and chooses the Normalized Discounted Cumulative Gain(NDCG)to evaluate the efficacy of the contrast model and the experimental model in this thesis.The experimental results show that the proposed method enhances the accuracy of tie strength prediction.Further,this thesis also proves the social network user's position in the same relationship is not equal,and the perception of the strength,from the two sides of interaction,is also not consistent.Therefore,it can distinguish the influence of the social network users effectively,and is beneficial to the research of opinion leader discovery and the information dissemination mechanism.