Early Prediction and Empirical Research of Network Public Opinion Information Heat Based on Deep Learning
[Purpose/significance]Social network platforms have become an important place for public opinion,and hot information spreads quickly and widely.How to predict the popularity of information in the early stage of information release has important re-search value in the era of big data,and has practical significance for the government to properly respond to public opinion events.[Method/process]Based on Weibo,this paper proposes a rating system for public opinion popularity,and takes"Zhengzhou trans-ported 870 people to Xuzhou without communication"as an example,uses a variety of neural networks to conduct prediction experi-ments,and fully verifies the index system.feasibility.On this basis,the ablation experiment is carried out to explore which index has the greatest impact on the prediction of public opinion heat.[Result/conclusion]Experiments show that the training and fitting state of various models based on this index system is good,among which the correct rate of Bi-GRU model is the highest,up to 92.41％.It shows that the index system proposed in this paper is reasonable and can provide a reference for the follow-up prediction of the popu-larity of public opinion.[Innovation/limitation]The theoretical basis for rating the popularity of online public opinion needs to be im-proved,and how to predict the popularity by integrating the features of multimedia information extraction is a direction that can be im-proved in future research.
internet public opinionpopularity predictionneural networkdeep learningearly prediction
NETL
NSTL
万方数据
