Abstract: Public opinion is an aggregate of people’s views, attitudes, and emotions about events that can spread through the Internet to generate online public opinion. Studying the evolution of online public opinion during emergencies can help relevant departments to take targeted measures to respond in advance. Tweets and Weibo texts with negative emotions are essential factors affecting the evolution of online public opinion. To this end, this paper proposes a novel improved grey model, SISGM(1,1), that optimizes initial conditions and background values for predicting the number of negative Weibo texts generated during emergencies. The model is improved as follows: First, the background value is reconstructed by the Simpson rule to achieve the effect of smoothing the data sequence. Second, the ISRU activation function is used to modify the initial condition, which can better reveal the characteristics of data growth and improve the model’s adaptability. Then, the modified background value is combined with the optimized initial condition to realize the double optimization. Finally, the PSO algorithm is used to calculate the introduced parameters to improve the prediction accuracy further. Additionally, the model is compared with five competing models to predict the evolution of online public opinion during emergencies. The experimental results demonstrate that the proposed model has apparent advantages compared to the other five competing models.

Key words: Emergencies, Online Public Opinion, Grey Model, ISRU Function, Simpson Formula