As a new carrier of public opinion and a novel approach for information propagation, the microblog plays a more and more important role in the initiation and spreading of the online public opinion. Compared with the traditional media such as blogs, forums and product reviews, the microblog has many new outstanding features, such as convenience, limited length and real-time. Based on these characteristics, in this thesis the model of influence propagation in Chinese microblog community is studied and the method of opinion leader mining is proposed based on the influence model.Firstly, due to large numbers of microblog users and the real-time information dissemination, it is difficult for the traditional model to characterize the process of the influence propagation in the social network constructed by microblog. In this thesis, the basic model and the discrete time model and the continuous time model are proposed to model the influence propagation process in microblog. The influence between users is detected by the activities that performed by the users. The experiments in the real world microblog dataset have validated that the proposed models can effectively reflect the influence propagation between users in microblog, and predict the users’ activities.Secondly, to analyze the content similarity between microblog texts, a method of similarity measurement based on the Chinese microblog syntax and semantic structures is proposed. Furthermore, a microblog short text classification algorithm is implemented on the basis of the proposed similarity measurement.Finally, based on the proposed model and methods, an improved opinion leader mining algorithm is presented. The opinion leaders in Chinese microblog social network are detected by mining the most influential microblog posts and users. Experiment results show that the proposed opinion leader mining algorithm can not only effectively discover the opinion leaders and their impact areas, but also can achieve a better performance. |