IPSO-LSTM hybrid model for predicting online public opinion trends in emergencies

• Published in: PloS one Vol. 18; no. 10; p. e0292677
• Main Authors: Mu, Guangyu, Liao, Zehan, Li, Jiaxue, Qin, Nini, Yang, Ziye
• Format: Journal Article
• Language: English
• Published: San Francisco Public Library of Science 10.10.2023; Public Library of Science (PLoS)
• Subjects: Algorithms; Analysis; Biology and Life Sciences; Computer and Information Sciences; Emergencies; Emergency management; Emotions; Evaluation; Forecasting models; Forecasts and trends; Long short-term memory; Management; Mathematical models; Mathematical optimization; Modelling; Neural networks; Particle swarm optimization; Physical Sciences; Prediction models; Public opinion; Research and Analysis Methods; Sentiment analysis; Social Sciences; Trends; China
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0292677

When emergencies are widely discussed and shared, it may lead to conflicting opinions and negative emotions among internet users. Accurately predicting sudden network public opinion events is of great importance. Therefore, this paper constructs a hybrid forecasting model to solve this problem. First, this model introduces an improved inertia weight and an adaptive variation operation to enhance the Particle Swarm Optimization (PSO) algorithm. Then, the improved PSO (IPSO) algorithm optimizes the parameters of the Long Short-Term Memory (LSTM) neural network. Finally, the IPSO-LSTM hybrid prediction model is constructed to forecast and analyze emergency public opinion dissemination trends. The experimental outcomes indicate that the IPSO-LSTM model surpasses others and has high prediction accuracy. In the four emergency predictions we select, the MAPE value of IPSO-LSTM is 74.27% better than that of BP, 33.96% better than that of LSTM, and 13.59% better than that of PSO-LSTM on average. This study aims to assist authorities in quickly identifying potential public opinion crises, developing effective strategies, and promoting sustainable and positive growth in the network environment.