Research On Source Tracing Of Public Opinion On Micro-blogs Combined With Impact Analysis

Abstract:The main purpose of source tracing of public opinion is to find the source and propagation way of public opinion. It is the basics of the information propagation study. Because the modes of information’s generation and propagation are not same in different network platforms, the methods of source tracing of public opinion are not same. As is known to all, micro-blogs are not only a famous social network but also have strong representation in public opinion spreading. Therefore, we set micro-blogs as the research object.The thesis firstly points out the characteristics of micro-blogs micro-blogging, such as short text and complicated theme. Then according to these characteristics and requirements for source tracing of public opinion, this paper presents a feasible method combining topic clustering with impact analysis.In order to tap the contact between different micro-blog’s content, this paper replaces the conventional method(word statistics), which is fit for long text, with improved word similarity algorithm in the process of micro-blogs similarity calculation. This algorithm is combined with depth attribute to solve the question that the specific degree of sememes meaning are different. All these works make the similarity between words more accurate, so that the correct rate of micro-blogs clustering is improved.This paper uses the impact analysis to filter out the micro-blog play the important role in public opinion propagation. Those micro-blogs are the main propagation paths of public opinion. After further analyze the essence of the impact, the micro-blog impact model which is compose of directly impact and subconscious impact is established. This model combines the various properties of micro-blog, such as retweeting number, comments number, user characteristics and so on, which can accurately express the real micro-blog impact.Each part of the source tracing of public opinion is respectively simulated using JAVA language. With analyzing the simulation result, it is proved that algorithms proposed in this paper are all local optimum. The thesis using these algorithms simulates the whole process of source tracing of public opinion and obtains acceptable result.