Mining Opinion Leaders Of Microblog Based On Social Network Models And Topic Evolution

Microblog, a virtual social network, provides a platform in the modern society for people to express their opinions, to get access to information and to discuss topics. This platform owns a great user base and the opinions arisen also have a significant influence on the real life. Microblog opinion leaders, influential both for their opinions and their hidden topics, have direct or indirect implication on many microblog users within a short time, or even control the whole progress. Therefore, with those opinion leaders, it would be more predictable for the whole picture or better to guide public opinion or even possible of pre-warning However, the existing excavating methods only consider the basic properties of social networks, somehow ignore the influence variation of those leaders involved according to the topic of evolution. In this regard, this paper mainly analyzes the microblog opinion leaders in the microblog social network varying along with the topic evolution, and researches on the tracking method.First of all, a graph model is built based on the microblog social network, in which nodes in the graph represent users while sidelines represent the relationships between the users (such as a single forward, multiple forwarding, number of comments, concerns, etc.). Through the users weights, the graph model is fulfilled to differentiate the weight degree, so that the structure can spot the potential opinion leaders and serve as the basis of further research.Secondly, on the basics of communication science, a new target-data segmentation model is built. Different from the previous graph model, this one uses the target-data partition weight as a vector to find three types of users through the similar features, then calculates the potential opinion leaders. Then those calculation results return to verify the effectiveness and accuracy of the model. Since differences still exist between the model and the link graph model, this results are not involved in the subsequent model calculation.Thirdly, based on the previous research, mining method for opinion leaders in the actual social network topics evolution is proposed. Based on the potential opinion leaders, with the model we can calculate the potential evolution of opinion leaders and the actual changes in public opinion, yet implement method based on topic evolution of opinion leaders. Compared to the algorithm only considering network model, this paper also considers the views of the opinion leaders in the actual public discussion. The experimental results show that, in the actual public opinion evolution, the found opinion leaders possess a more obvious representative.The results are tested on the currently most popular microblog social network-Sina Weibo, and show that the opinion leaders discovery modal based on the topic evolution can find better and more relevant opinion leaders related to the evolution degree in the whole public opinion, namely that this model can find public opinion which is practical and relevant to the topic evolution.