Analysis of Public Opinion on the Network of Burst Events based on the PH-LSTM Neural Network

Abstract

Due to the high-speed development of the Internet today, once a burst event occurs, it will attract the general public‘s attention in a short time and spread rapidly on the Internet, which will lead to the formation of large-scale network public opinion on emergencies. We can accurately judge the development trend of bursts and provide a reference for emergency supervisory departments to deal with public opinion crises by predicting the trend of public opinion on the network of emergencies. To address the issues of a single prediction model's low prediction accuracy and the large influence of social media factors on public opinion trend, an opinion prediction method integrating social media influence and LSTM neural network is proposed, and a PH-LSTM (Post Hot-Long Short-Term Memory) neural network is built. Firstly, the influencing factors of online public opinion are introduced, including the posting hotness and user influence of fused social media influence and the growth rate of events. Secondly, the LSTM neural network is improved to design a PH-LSTM prediction model with three hidden layers, and the PH-LSTM model is used in the quantitative prediction of unexpected events. Model’s validity was verified through experiments, and it was confirmed that the model has a good fitting and prediction effect on the trend of online public opinion of unexpected events.