| In recent years，the network has become the main platform for the pubic toexpress personal feelings and political attitudes，the formation and disseminationof mass unexpected incidents network public opinion has also had a major impacton netizen and society. As the network enriched in the presence, especially the rapidgrowth of Micro-blog, the breed and spread channels of mass unexpected incidentshave shifted, Micro-blog has already become the main carrier of network publicopinion，and the newly formed Micro-blog public opinion presented differentcharacteristics in evolution process and evolution manner.This paper completed a whole analysis on the evolution process of the massunexpected incidents Micro-blog public opinion, including the static evolutionparameter studies, and the dynamic evolution aspects analysis.First, the raw research data was collected from Sina Weibo platform, and theamount of micro-blogs, as well as the frequency of reposts, were used as dominantindicators to reflect the overall heat condition of Micro-blog public opinion; micro-blog texts were clustered into different themes by the integrated use of probabilistictopic model and text clustering method based on distance; by comparing theclassification performance of different feature extraction algorithms and textclassification method portfolios, a text classification model was built with TF-IDF,IG and SVM, with which the sentiment polarity classification of micro-blog textswas completed precisely. Then, this paper analyzed the evolution condition of allthe above parameters, as well as their performed characteristics in eachevolutionary stage. Finally, by taking the survival time and reposting times as maindependent variable, this paper constructed an opinion leaders identifying modelwith Cox’s proportional hazard analysis and Negative Binomial regression analysis,and the effects of the identified opinion leaders to Micro-blog public opinionevolution were studied subsequently. |
PDF Full Text Request