Research On The Mechanism Of Following-based Topics Of Microblogging Social Network

As one of the most popular social applications in Web2.0era, microbloggingdevelops rapidly in recent years. Its unique characteristics like simplicity, real-time andconvenience makes the microblogging growing exponentially and the dissemination ofinformation more and more convenient. However, in some vertical applications, users cannot implement valid concern, thus affecting the universal application of microblogging inthese areas. Research on the mechanism of following based on topics of microbloggingsocial network has a good value in practice.The research on the mechanism of following based on topics of microblogging socialnetwork has provided a complete set of analysis mechanism to solve the problem offollowing the core members and new users’"cold start". Firstly, when facing the massmicroblogging data in Internet, it proposes a method based on Sina SDK to obtainmicroblogging data, provide data sources for microblogging social network; Secondly, onconsidering of the semantic information of microblogging is generally dificient, we haveproposed a method for classification of microblogging short text. The method works byusing the HowNet ontology library to extend its semantic information and using the KNNclassification algorithm based on users’ feedback to improve the classification accuracysignificantly. Furthermore, based on the comparison of different network model, we haveproposed a strategy to construct the user interaction network based on the topics. Finally, byanalyzing the microblogging network topology characteristics, we have designed animproved algorithm MutualRank, which used for discovery the core members ofmicroblogging based on the PageRank algorithm of web link network.The experimental results have showed that the acquisition methods for microblogginghave gained a rich set of data resources, and also have a good text classification results:performance is improved about25%. By analyzing and comparing the interactive networkof microblogging based topics we have built, we have found that the MutualRank algorithmis more suitable for the microblogging network in the discovery of core members, and it haslaid a solid foundation for the following personalized recommendation.