| 基于多主体博弈与深度强化学习的舆情治理策略优化模型 |
| Strategy Optimization Model for Public Opinion Governance Based on Multi-Agent Game and Deep Reinforcement Learning |
| 摘要点击 40 全文点击 0 投稿时间：2025-04-25 修订日期：2026-06-11 |
| 查看/发表评论 下载PDF阅读器 |
| 中文关键词 多主体博弈; 深度强化学习; 舆情治理; 策略优化; 演化博弈 |
| 英文关键词 multi-agent game; deep reinforcement learning; public opinion governance; strategy optimization; evolutionary game |
| 基金项目 安徽省哲学社会科学青年项目 (AHSKYQ2024D016) |
| 投稿方向 舆情治理；演化博弈 |
| 作者 | 单位 | 邮编 | | 吕金辉* | 淮南师范学院 | 232038 | | 郭芳芳 | 淮南师范学院 | |
|
| 中文摘要 |
| 网络舆情的迅速传播和演化给社会治理带来重大挑战. 本文融合多主体博弈理论与
深度强化学习方法，提出舆情治理策略优化模型. 首先，构建政府监管者与传播者的多主体博
弈模型，分析模型的纳什均衡条件，严格证明了存在惩罚阈值使谣言传播状态发生突变的定
理. 其次，设计深度强化学习算法求解最优治理策略，将博弈理论分析融入多智能体强化学
习框架，证明算法在特定条件下收敛于纳什均衡策略. 通过仿真实验，结果表明：政府惩罚力
度必须超过临界值才能有效抑制谣言传播，这与理论分析一致；相比固定策略，本文模型通过
深度强化学习自适应调整治理策略，明显降低舆情传播规模和速度；模型对参数扰动具有较
强鲁棒性. 本研究为网络舆情治理提供了一种将博弈论洞察与智能决策相结合的创新理论框
架，有助于优化复杂舆情环境下的治理策略. |
| 英文摘要 |
| The rapid spread and evolution of online public opinion poses significant challenges to social governance. This paper integrates multi-agent game theory with deep reinforcement learning methods to propose a strategy optimization model for public opinion governance. First,
a multi-agent game model between government regulators and propagators is constructed, analyzing the Nash equilibrium conditions and rigorously proving the theorem that there exists a punishment threshold causing a phase transition in rumor propagation states. Second, a deep reinforcement learning algorithm is designed to solve for optimal governance strategies, integrating game theory analysis into a multi-agent reinforcement learning framework, proving that the algorithm converges to Nash equilibrium strategies under specific conditions. Through simulation experiments, the results show that: government punishment intensity must exceed a critical value to effectively suppress rumor propagation, consistent with theoretical analysis; compared to
fixed strategies, this model adaptively adjusts governance strategies through deep reinforcement
learning, significantly reducing the scale and speed of public opinion propagation; the model
demonstrates strong robustness to parameter perturbations. This research provides an innovative theoretical framework for online public opinion governance that combines game theory
insights with intelligent decision-making, helping to optimize governance strategies in complex
public opinion environments. |
| 关闭 |
