Research On The Linked Data Mining In The Recognition Of Microblog Opinion Leaders

At present,social networking is one of the main ways to get information and maintain interaction.The massive data are rich in the use value.Accurate and efficient identification of opinion leaders in social networks is important for guiding public opinion,business investment,and avoiding risks.But the heterogeneity of social network data and the complexity of the relationship have limited the accuracy of the research and the efficiency of the algorithm.Using the traditional data as the research object will miss a lot of hidden semantic information,but the linked data can be a good solution to these problems,accurately express the meaning of the data,dig out more effective information,also for the follow-up research data sharing and Extensions are facilitated.In this thesis,we calculate the similarity of microblogging users to lay the foundation for the realization of microblogging data association,and combine the constructed data with association rules mining algorithm and index scoring method to solve the problem of objectivity recognition.The research work of this thesis is divided into three aspects:Aiming at the problem that the accuracy of the traditional similarity calculation method is not high because of the chosen attribute,a comprehensive similarity calculation method based on microblog data is proposed.By analyzing the user’s attributes,select two parts of information—background and interactive information as the standard,choose the calculation method according to the specific data structure of the corresponding,use the statistical information and AHP(analytic hierarchy process)to determine attribute values,calculate the similarity between users.Use accurate Rate,recall rate,F1 measure as the evaluation index of the experimental results.The experimental results show that the comprehensive similarity calculation method can be more accurate to measure similarity between users.Combined with the attribute analysis and the results of the comprehensive similarity calculation method,learn from the idea of ‘Seven-step’ ontology construction method,use protégé to build the microblog linked data by defining classes,defining properties,and creating instances.Aiming at the problem of subjectivity in selecting attributes and lack of accuracy of traditional method,a method based on linked data is proposed.By using the depth-first algorithm to process the data and express the information completely,using the Apriori algorithm to generate the association rules related to the opinion leaders,and determining the measurement factors objectively,combined with the index scoring method,identify opinion leaders.The experimental results show that the feasibility of the recognition method based on the linked data.