Research On Several Types Of Network Public Opinion Modeling And Application Based On Big Data

The rapid development of big data and internet has promoted network public opinion from theoretical analysis to applied analysis.Based on the existing research results of online public opinion at home and abroad and my own preliminary research work,this thesis uses big data technologies and methods such as crawler technology,correlation analysis,visualization analysis and combination prediction to carry out application research on internet public opinion in the three fields of overseas tourism,colleges and universities negative public opinion,and derivative public opinion.The research on overseas tourism has built a new set of public opinion indicators system,using 4 first-level indicators and further subdivided 9 second-level indicators to quantitatively analyze the factors affecting overseas tourism and public opinion,and get the numerical ranking of the correlation degree of different public opinion indicators,giving the tourism industry to provide some thinking direction under the influence of the epidemic.The research on negative public opinion events in colleges and universities adopts a variety of research methods from 4 different perspectives,and visualizes the relevant online public opinion information with negative impacts and complex data,proposes special solutions for different perspectives,and analyzes the same type of online public opinion events.The result can promote and use in conjunction with specific events.The research on derived public opinion topics is mainly based on data information,supplemented by text information,and combined predictive modeling ideas and clustering algorithms are used to achieve hierarchical early warning of derived public opinion.The combined forecasting method is a combination of the Logistic model,ARIMA model and Holt-exponential smoothing model with different weights obtained by gray correlation analysis,and the results of the combined model need to be tested and passed.The clustering algorithm is used to calculate the grading of the derivative coefficients,and then divide the early warning levels of derivative public opinion events into three levels: mild,moderate and severe.Finally,the validity of the model is proved by a verification case,and put forward my own thinking about the derivative public opinion analysis of emergencies.