| Internet applications such as social networking have accelerated the speed of information sharing and dissemination among people.However,the convenience,concealment,and low-cost characteristics of online platforms can easily breed false news,cause public opinion storms,and increase people’s anxiety.Therefore,effective early warning analysis of Internet public opinion can not only gain more response time to take countermeasures,but also further compress the information vacuum,prevent the spread of false news,and reduce the negative impact of the public opinion.A detailed analysis and summary of the early warning schemes in the monitoring of Internet public opinion is presented.According to different early warning models used in different application scenarios,public opinion early warning models are classified into three categories: models based on time series,models based on artificial intelligence,and models based on combination optimization.In-depth comparative analysis of representative models is explored,the advantages and disadvantages of the Internet public opinion early warning model as well as the existing challenges are summarized,and the future research directions and development trends are analyzed.An early warning RVM-L-LIM model combined with Lagrange interpolation method is proposed(Relevance Vector Machine,Logistic,Lagrange’s Interpolation Method,RVM-L-LIM).Focusing on the issue of rapid data accumulation,multiple factors and insufficient data in the public opinion trend prediction.Considering that the current public opinion is affected by multiple factors,the RVM-L is used to obtain a comprehensive heat value to ensure the compatibility of public opinion heat indicators.When the amount of data is insufficient,the LIM method is adopted to predict the data in the valid interval based on the historical data to fill in the amount of historical data for model prediction.Finally,the trend correction is carried out on the forecast data,and optimize the trend forecast result of public opinion.An Internet early warning mechanism based on the critical zone is proposed.In the early warning analysis stage of public opinion,the trend of public opinion dynamic changes and it’s difficult to obtain the early warning points accurately.Based on the forecast trend of RVM-L-LIM model,the key points of the trend change are calculated;the critical area at the early warning point is determined according to the key points and the error variation range,and the critical area is divided into different levels;early warning interventions are implemented in each critical area to analyze the trend of public opinion.On the real data set of Sina Weibo,the effectiveness of the propose is verified.The predictive performance of different early warning models under different data states is analyzed,in terms of correlation coefficient(R-square)and root mean square error(RMSE)indicators.The intervention performance in different critical regions is analyzed.The peak changes with the intervention and the impact of the key points of the trend on public opinion are studied in detail.The experimental results show that the RVM-L-LIM model can achieve the prediction accuracy of 96%,and the early warning in the critical region can effectively control the trend of public opinion and reduce 60% peak value. |
