突发敏感公共事件背景下网络舆情信息评估模型构建——基于熵权-AHP组合赋权与TOPSIS-灰色关联混合评估方法

Construction of a Network Public Opinion Information Assessment Model in the Context of Sudden Sensitive Public Events—Based on Entropy-AHP Combined Weighting and TOPSIS-GRA Hybrid Assessment

摘要: 突发敏感公共事件网络舆情具有突发性高、传播速度快、社会敏感性强等突出特征，构建科学有效的舆情信息评估模型对于公共管理部门精准防控舆情风险具有重要理论与实践价值。本文遵循科学性、系统性、可操作性原则，构建涵盖舆情传播热度、舆情内容敏感度、舆情主体影响力、舆情演化态势4个一级指标、12个二级指标、35个三级指标的网络舆情信息评估指标体系，创新性引入话题敏感等级、公众权益关联度、多部门协同需求度等特色指标。运用灰色统计法筛选保留28个核心指标，采用熵权法与层次分析法(AHP)进行组合赋权，融合逼近理想解排序法(TOPSIS)与灰色关联分析法(GRA)构建混合评估模型。以5起典型突发敏感公共事件为样本的实证分析表明，所构建的混合评估模型能够有效识别舆情风险等级，评估结果与实际发展态势高度吻合，显著优于单一TOPSIS法和单一GRA法；舆情内容敏感度和舆情传播热度是影响风险评估结果的关键维度；模型在赋权方法变更、参数扰动和样本增减等多重检验中保持稳健。本文将TOPSIS-GRA混合模型引入突发敏感公共事件舆情评估领域，构建了融合公共管理领域独有指标的评估体系。

Abstract: Network public opinion on sudden sensitive public events is characterized by high abruptness, rapid dissemination, and strong social sensitivity. Constructing a scientific and effective public opinion information assessment model holds significant theoretical and practical value for public administration authorities in precisely preventing and controlling public opinion risks. Following the principles of scientificity, systematicity, and operability, this study establishes a network public opinion information assessment index system comprising 4 primary indicators (public opinion dissemination heat, content sensitivity, subject influence, and evolution trend), 12 secondary indicators, and 35 tertiary indicators. Innovative characteristic indicators—including topic sensitivity level, public interest relevance degree, and multi-department coordination demand—are introduced. The Grey Statistical Method is employed to screen and retain 28 core indicators. The Entropy Weight Method and Analytic Hierarchy Process (AHP) are combined for weighting, while the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) and Grey Relational Analysis (GRA) are integrated to construct a hybrid assessment model that simultaneously captures both positional proximity and shape similarity. Empirical analysis of five typical sudden sensitive public events demonstrates that the proposed hybrid assessment model effectively identifies public opinion risk levels, with assessment results highly consistent with expert judgments, significantly outperforming the single TOPSIS method and the single GRA method. Public opinion content sensitivity and dissemination heat are identified as the key dimensions influencing risk assessment. The model exhibits sound robustness under multiple tests including weighting method substitution, parameter perturbation, and sample addition/removal. This study is the first to introduce the TOPSIS-GRA hybrid model into the assessment of public opinion on sudden sensitive public events, establishing an evaluation system incorporating unique indicators from the public administration domain.