Research On Online Public Opinion Heat Prediction Of Public Health Emergencies Based On Sentiment Analysis

Purposes:Using the sentiment analysis method integrating emoticon feature of online public opinion obtains sentiment tendency,and add the sentimental factor to heat prediction basis of online public opinion to realize the research on online public opinion heat prediction of public health emergencies.Methods:The literature research method is used to summarize the current research status of online public opinion prediction and online public opinion sentiment analysis at home and abroad.Based on current research status,an online public opinion sentiment analysis method incorporating emoticon feature and an online public opinion heat prediction model for public health emergencies based on sentiment analysis are proposed.In the empirical study,the public health emergency of “×××vaccine fraud” was used as the research object.After preprocessing the data,Word2 vec was used to supplement new words in the field,then Doc2 vec and SVM were used to realize the sentiment analysis of online public opinion,and finally used LSTM carries out multivariate time series analysis to realize the research of online public opinion heat prediction.Results:The multi-feature fusion sentiment analysis method that integrates dictionary feature,emoticon feature and vector feature has the best effect,and its recall rate,accuracy rate,F1 value and AUC reach 85.5%,85.2%,85.0% and 0.91 respectively.When the lag is 5 days,online public opinion heat prediction model with sentiment value is the best,and its RMSE is 223.089 and MAPE is 0.150.Conclusion:Taking emoticon feature as one of the feature types of sentiment analysis and integrating it into dictionary feature and vector feature can improve the sentiment analysis effect of online public sentiment.The effect of the online public opinion heat prediction model with sentiment value is better than that of the online public opinion heat prediction model without sentiment value.Introducing emotional factor into the online public opinion heat prediction of public health emergencies can improve its prediction effect.