Research On Topic Heat Prediction Based On VOLDA Theme Model And ESG Prediction Model

With the popularization of network,the Internet and Web2.0 social media have a great impact on people's daily life and produced a huge amount of texts of online public opinion.These online public opinion texts such as forum posts contains of rich topic content information and can real-time reflect the current hot topic of network public opinion.In addition,after the analysis of the online public opinion texts,we can follow the hot topics that people pay attention to.Therefore,it is of great significance to do online topic modeling for public opinion text and to excavate topic information in real time and to calculate and predict the heat of topic.However,the topic composition of public opinion text is ignored in the heat calculation of existing topics,and the prediction ability of existing prediction models is not good enough for topic fever,so this paper proposes a topic heat prediction method based on the VOLDA theme model and the ESG prediction model.This paper first studies the related theories of online topic model and time series forecasting method as the research basis.Secondly,the theme similarity matrix is introduced into the OLDA model to build a variable online LDA model(Variable Online-LDA)to remove the impact of the unrelated time slice theme.In addition,considering that time series is usually composed of multiple time series modules,the ESG(EEMD-SVM-GMDH)prediction model is proposed for time series decomposition prediction in this paper.After that,we designed a topic heat prediction method based on the above two models,and used VOLDA model to automatically get topic heat time series,and used ESG model to predict it.Finally,this paper crawls the Tianya forum post to build experimental data set for experiment.The experiment proves that the model and method proposed in this paper are effective and reliable.The specific innovation points are as follows:(1)This paper constructs the VOLDA theme model and the ESG prediction model.The VOLDA model removes the current theme independent time slices in the theme content evolution matrix to avoid the influence of unrelated topics and improve the effects.The ESG prediction model decomposes the time series,and adds the weighted prediction results to get the final prediction value,which makes full use of the original time series information and improves the prediction effect.(2)This paper puts forward the prediction method based on VOLDA and ESG.It combines topic information and external heat characteristics define topic fever more reasonably,and automatically acquire the thermal time series in the process of VOLDA theme modeling,and use the ESG model to decompose the prediction to achieve better prediction results.