关键词: 群体观点, 突变论, 弹性, 文本分析, Q学习, 粒子群算法

Abstract: Sudden bifurcations and reversals frequently occur in the evolution of online group opinions, posing significant challenges for governance by both governments and enterprises. As a complex system, online group opinion dynamics are difficult to explain and regulate effectively using traditional opinion dynamics models, system modeling, or control methods. To address this issue, this paper extracts group opinions and their influencing factors from online forums through text analysis, and constructs a cusp catastrophe model to represent opinion dynamics near critical bifurcation points. Additionally, it establishes a resilience index model to quantify the internal accumulation of pressure caused by external shocks, thereby enabling the monitoring of potential opinion mutations. Furthermore, it proposes a Q-learning-enhanced particle swarm optimization (PSO) algorithm to regulate the resilience index and maintain the stability of group opinions. Empirical validation using data from the “Meituan Delivery Forum” demonstrates the effectiveness of the proposed method and uncovers the regulatory patterns of group opinion control. This paper contributes a methodological innovation in controlling nonlinear behaviors in complex systems by integrating catastrophe theory, resilience analysis, and intelligent algorithms. It also provides a framework for identifying key control factors in managing sudden opinion shifts, offering decision-making support for social organization managers.

Key words: group opinion, catastrophe theory, resilience, text mining, Q-learning, particle swarm optimization (PSO)