Dynamic Opinion Tree Research Of Microblog Hot Topics Based On Evolutionary Clustering

As a new public media, micro-blog has attracted most of netizens and become an important platform for the generation and propagation of hot topics. However, the rapid development of the micro-blog, making the risk of unexpected events increased, all of which has brought great difficulties for related orgnizations to deal with the emergencies and social crisis measures. Therefore, the opinion mining research on hot topics can help government users to quickly understand and grasp the demands and features of netizens’ view.Based on the research of scholars, combined with the characteristics and life cycle of hot topics, this paper made definition of the concept of dynamic opinion tree and analyzed the process of building this tree. Secondly, for the lack of vector space model(VSM), this paper constructed an sentiment vector space model(S_VSM), combined with the principles of PMI-IR and VSM. Thirdly, the type of micro-blog hot topics is dynamic, which is inadequate for the traditional clustering algorithm. This paper proposed the concept of evolutive k-medoids algorithm flow firstly and had done a series experiment to prove its stability. Finally, based on the evolutive k-medoids algorithm, this paper built an dynamic opinion tree, using the real-world data of micro-blog hot topics, and analyzed the characteristics of netizens’ view and made recommendations for management for the government.The contribution of this paper is mainly reflected on:Firstly, this paper made definition of the concept of dynamic opinion tree and analyzed the process of building this tree, and introduced hierarchical structure detailedly and showed its significance in the management of the public opinion and sentiment.Secondly, f this paper constructed an sentiment vector space model(S_VSM),based on the principles of PMI-IR and VSM. for each micro-blog comment S_VSM could generate a three-dimensional vector for sentiment analysis and opinion clustering.Thirdly, This paper proposed the concept of evolutive k-medoids algorithm flow firstly and built an dynamic opinion tree, using the real-world data of micro-blog hot topics, and done a series experiment to prove its stability.