融合时序状态估计与非线性社区检测的社交网络回声室检测模型

An Echo Chamber Detection Model for Social Networks Integrating Temporal State Estimation and Nonlinear Community Detection

摘要: 回声室检测工作是舆情防控中的基础任务。针对现有的回声室检测模型在考虑社区结构、节点属性、信息传播过程和时间因素时存在的不足，影响了节点间互动频率、信息传播概率、信息流动方向等特征的准确性的问题，本文提出了一种基于时序状态估计的社区回声室检测模型。利用改进的无迹卡尔曼滤波对社交网络进行状态估计，并设计融合节点极性和社区极性影响的传播模型。实验结果表明，与当前主流的滤波算法进行比较，该模型在社区中的回声室检测任务性能表现更好，证明所提出的模型能有效提升回声室检测的有效性，且在舆情管控中具有良好的应用价值。

Abstract: Echo chamber detection is fundamental in public opinion control. Addressing the limitations of existing models concerning the consideration of community structure, node attributes, information dissemination processes, and temporal factors, which affect the accuracy of features such as interactivity frequency, information propagation probability, and information flow direction, this paper proposes a state estimation-based model for community echo chamber detection. Utilizing an improved Unscented Kalman Filter for state estimation of social networks and designing a propagation model that incorporates node polarity and community polarity influences, the model outperforms mainstream filtering algorithms in echo chamber detection tasks within communities. The experimental results demonstrate the effectiveness of the proposed model in enhancing echo chamber detection and its significant applicability in public opinion control.