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)

摘要： 在线群体观点演化中的突然分岔与反转等现象频发，已成为政府与企业共同关注的治理难题。在线群体观点演化作为一个复杂系统，传统的观点动力学、系统建模及控制方法难以解释其突变的内在机制并实现有效调控。为此，提出一种创新方法：通过文本分析提取在线论坛中的群体观点及其影响因素，构建群体观点处于突变临界区的尖点突变模型；建立群体观点的弹性指数模型，用于测量群体观点受外界冲击时产生的内部压力积累，并以此监测群体观点突变的可能性。进一步，提出融合Q学习的粒子群优化算法，通过控制弹性指数来维持群体观点稳定性。以“美团配送吧”的实证数据分析验证了所提方法的有效性，揭示了群体观点控制的关键规律。本文在复杂系统非线性行为控制方面实现了方法论创新，融合了突变理论、弹性概念与智能算法，形成多维度突破。为在线群体观点突变的控制提供了关键控制因素识别方法，同时为社会组织管理者提供了决策支持。

关键词: 群体观点, 突变论, 弹性, 文本分析, Q学习, 粒子群算法