Prediction of complex public opinion evolution based on improved multi-objective grey wolf optimizer

Abstract

In an emergency, predicting the evolution trends of network public opinion is of great significance for governments to manage the situation and associated social public opinion trends. Most of the existing public opinion prediction models have the disadvantages of long time consumption, limited types of predicted public opinion categories and low-accuracy prediction results. To improve the ability to predict the evolutionary trend of complex public opinion, an improved multiobjective gray wolf optimizer (IMOGWO) is proposed. IMOGWO uses logistic and Lotka-Volterra models to initialize the wolf population and improve the population validity; it designs a nonlinear function to adjust the population update factor to improve the exploration and exploitation ability and the local search ability of the wolf population; and it introduces the elite retention policy and a Pareto-optimal solution set to achieve multiple objectives using the principle of a nondominated solution set. In simulation experiments, various network emergencies were taken as empirical analysis cases. The experimental results show that the improved multiobjective GWO has good accuracy and universality on complex public opinion evolution prediction and can better predict the evolution trends of various types of complex public opinion than other tested methods.