Research On Microblog Public Opinion Monitoring System Based On Cloud Computing

Recently, the method of finding topic from microblog and opinion monitoring analysis of these topics were practical significant, because many social hot topics often begin with microblog. This article study the implementation of the microblog public opinion analytics based on the characteristics of microblog. The advantages of Hadoop platform in data processing, data collection technology, text modeling, clustering algorithm and realization of public opinion function were introduced in this paper. The main work is as follows:Firstly, in this paper, microblog data is collected by microblog open API interface and web crawler. The achievement of data acquisition is based on the search of the hot search terms. At the same time, we get a large number of microblog data in a short period of time.Secondly, in the theme discovery of microblog text, the discovery of new words is proposed to improve the accuracy of topic discovery, which is divided into three steps: segmentation for Chinese words, LDA theme model, theme clustering. Segmentation accuracy of this article is improved by the new word detection module in Chinese word segmentation. In the establishment of the text model, the accuracy of the topic discovery is improved by the LDA theme model.Thirdly, in order to improve the accuracy of LDA modeling, the text classification module is added into the model, which solve the problem of the deficiencies in the topic detection and different topic with the same keyword.This paper tests the function of the microblog public opinion monitoring system, including hot topic detection, topic diffusion path, contents and general trend and social network analysis, verifies the microblog public opinion monitoring analysis system which has achieved the target by analyzing the simulation results. Finally, for the defects of the system, future work is described.