Identifying Opinion Leaders In Social Network Based On User's Behaviors And Relationships

With the rapid development of communication techniques such as 3G,4G,HTML5,cloud computing and mobile internet,information spreads so quickly in the new era of mobile,digital,intelligent,and interactive.Social networks are characterized of personalized,interactive,virtual,social and user-centered.They have been widely applied in the system recommendation,social information security and knowledge sharing.Social networks have become one of the most influential platform of new media development.However,social networks face a series of problems in the field of system recommendation,social information security and knowledge sharing.Fortunately,opinion leaders in social networks can enrich the information of the network and play a significant role of filtering,controlling and bridging.Identifying and training opinion leaders of social networks is an important way to solve those problems.Therefore,identifying opinion leaders in social networks is a hot topic.Based on the intensive study on evolution path and related works of opinion leaders,this thesis summarized the definition and characteristics of the opinion leader in social networks considering user behaviors and relationships.After analyzing user behaviors and relationships,we did empirical research to demonstrate the existence of opinion leaders.This thesis put forward a set of systematic identification indexes of opinion leaders in social networks based on user behaviors and relationships.The identification indexes based on user behaviors are activity and centrality,and the one based on user relationships is intermediary.User behaviors reflect the state of user behaviors in social networks,and user relationships reflect the relationship network formed by the interaction between users.Combining user behaviors and relationships can reflect user behaviors and characteristics comprehensively.Moreover,this thesis proposed the identifying model of opinion leaders based on the presented indexes.It includes 3 steps:identifying the activist according to activity,identifying the topic opinion leader in consideration of centrality,and identifying the truth opinion leader according to the intermediary.This thesis also designed a system framework of identifying opinion leaders in social networks based on user behaviors and relationships.The system framework is composed of five components,such as data acquirement,information extraction,diagram construction of social networks,identifying opinion leader and results visualization.We got and extract the data with user behaviors and relationships from SIN A microblog,then classified information subject using MALLET software,and identified the opinion leader using Python language.Compared with the opinion leaders identified using PageRank and HITS algorithms,the identified leaders using the model proposed in this thesis conforms to the reality better.The proposed identification indexes and model in this thesis consider not only the network structure and link analysis,but also user behaviors and post content.Moreover,the operation is flexible and the result is visual.