Research On Key Users Minging Of Online Social Network

In recent years,Facebook,Twitter,Weibo,WeChat and other social networking sites have gradually become major social channels.People build relationships through social media,resulting in different levels of virtual online social networks.Online social networks are composed of the users and their dynamic interactions,in order to support the exchange of real-time network information and data.Therefore,the key users mining research can help enterprises to promote products,government to control public opinion,it has great value.Taking realistic online social network data for the study,the paper mainly mine hot topics opinion leaders,persistent topic key person analysis,key users in information sources and engine node mining research,then apply on the false information control and on the efficiency network marketing.The contents of this dissertation include the following four parts:Firstly,the existing mining algorithm cannot find out the topic-specific opinion leaders,which limits the effect of the public opinion control.Traditional clustering method is difficult to accurately identify topic related posts and clustering them together and the application of text sentiment analysis is not considered,so it is hard to achieve the expected effect.This paper provides a BBS opinion leader mining algorithm based on topic model(TOLM).The study first preprocesses the post titles based on their publication date,and then does further analysis using a semantic model based on latent Dirichlet allocation(LDA)which combines with TF-IDF.In the end,the algorithm sets up variable scale posts reply relational network for social network analysis and sentiment analysis,and ranks the users' influence to indentify the opinion leader.The TOLM algorithm is designed to mine opinion leader in a network hot events quickly,and has higher practicability with considering topic attributes,sentiment orientation and network structure.Then feasibility and effectiveness of the model is verified by experiments so that this research may benefit to correctly grasp and guide the direction of the development of the public opinion propagation.Secondly,as a real-time public information platform,BBS topics are divided into sudden topics and persistent ones which are close to the life of the people's livelihood with a longer time span.The difficulties of the persistent topic key person mining are the topic extraction and key nodes mining in sparse network,so our research mainly focuses on the analysis of the key uses in the persistent topic.The research mainly include discovery of the persistent topics and extract the key person in the persistent topic social network(PTSN)through sentiment weighted node position(SWNP).Identifying persistent topic mainly combines the LDA model and similarity model on the timeline.SWNP is a new method of node position analysis,which takes into account both the node position of the neighbors and the strength and emotional tendency of connections between network nodes.The methods in real data sets validate the effectiveness.Thirdly,as a kind of typical social networking platform,microblog information transmission speed very rapidly,finding the true source of a social network is a crucial component of social network information tracing.In view of the existing algorithm only find the earliest time stamp nodes,without considering relationship between nodes of social network(friends,attention,etc.)and their semantic relationships.Using the new media microblog as an example,this chapter provides a source tracing algorithm ITEPE(Initiators and Early Participants Extraction)to solve this problem.First,the cascade(session tree)is built according to the retweeting of a microblog,after which the cascade set(session forest)is clustered by topical relevance.Second,real initiators are identified through the user relationship network and information cascade network.The influence index and conformity index of every node is then iteratively calculated according to text sentiment analysis and information cascades and the early important participants are extracted.Finally,the real initiators and early participants are evaluated through an experiment.The source tracing nodes and the global high influence nodes are deleted to control the false information.Experiment results validate the proposed scheme in a real sina microblog dataset.Compared to previous studies,the cascade algorithm can be forwarded by the relationship between the collection formed traceability information.Finally,Retweeting is the main propagation mechanism in microblog platform,and the information often "fission" spreads centered on "engine nodes".The paper provides engine nodes mining algorithm,first,the cascade(session tree)is built according to the retweeting of a microblog,after which the pruning strategy is used according to the timestamp.Second,the cascade set(session forest)is clustered by topical relevance.Then the engine nodes with different precision can be extracted through computing the integrated diffusion capacity.Finally,the method is evaluated through an experiment.Microblog marketing is a low cost high efficiency marketing tools.The experiment result shows that the engine nodes found using the proposed method can speed up the propagation of information.