Research On Weibo Public Opinion Real-time Monitoring System Based On Thermal Model

With the rapid development of Internet technology,a large number of online social platforms have appeared on the Internet.Weibo is one of the representative.Because of the timeliness,autonomy,and interactivity of Weibo,it can improve and break through the shortcomings of traditional media,so it has been greatly developed.It has become an important way for Internet users to publish and share information,and it has gradually evolved into a popular Internet.Public opinion platform.And it is precisely because of its more free and diverse way of speaking,the outbreak of public opinion events often shows a virus spread,which brings great challenges to public opinion monitoring.This thesis focuses on the real-time monitoring of Weibo public opinion,mainly focusing on the two major issues of sentiment analysis and user influence analysis.Firstly,for the problem of sentiment classification,an sentiment classification algorithm based on ensemble learning is proposed.In this algorithm,the microblog crawler is first used to crawl target related microblog text information to obtain the original data,and then the original data is segmented to remove stop words.After preprocessing,the TF-IDF method is used to extract feature vectors,and the SVD method is used to reduce the feature vectors.Then,the stacking integration strategy is used to set up five basic classifiers to form an emotion classification model to determine the emotion classification of the data.Then,for the problem of user influence analysis,an influence analysis algorithm based on PageRank and HITS is proposed.In this algorithm,the microblog crawler is first used to crawl the target microblog user relationship network,and the PageRank and HITS algorithms are used to calculate the complex network respectively.,Calculate the influence of the nodes,get the Page Rank,hub,and authority values to form a feature vector,and then use the RankSVM algorithm to sort the influence of the nodes to get the final user influence.Finally,through a variety of comparative experiments,the effectiveness of the proposed method is demonstrated.At the end of the article,the work of this paper is summarized and the next work is prospected.