
Public Opinion Risk Early Warning Model on Government-Citizen Interaction Data: A Perspective on Evidence-based Decision Making
Liman ZHANG, Yueting U, Wenjing CHENG, Tianyi LIU, Xinxin SUN
Public Opinion Risk Early Warning Model on Government-Citizen Interaction Data: A Perspective on Evidence-based Decision Making
[Purpose/Significance] The study aims to construct an early warning model of public opinion risks based on government-citizen interaction data, guided by evidence-based decision-making theory. We seek to uncover the governance value embedded in such interaction data, providing new insights and methods for identifying and managing potential public opinion risks. Traditional methods of monitoring public opinion often rely on subjective judgment, leading to potential bias and inefficiency. In contrast, this study uses objective, data-driven techniques to improve the accuracy and reliability of risk predictions. By integrating evidence-based decision making with public opinion analysis, the study not only advances the theoretical framework but also provides practical tools for government use. This innovation is significant as it addresses the gaps in the current literature regarding the objective assessment of public opinion risks and their impact on governance, thereby contributing to the field of public administration and social governance. [Method/Process] The research methodology involves a multi-step process, starting with the identification of key indicators of public opinion risks. These indicators include appeal purpose, text length, sensitivity, emotional tendency, and degree of aggregation. The analytical hierarchy process (AHP) and the criteria importance through intercriteria correlation (CRITIC) method were employed to calculate the weight of each indicator. AHP, a subjective weighting method, uses expert judgement to construct a judgement matrix and determine indicator weights. However, to reduce subjective bias, the CRITIC method is integrated, which objectively determines weights based on the variability and conflict in the data. The model's workflow began with problem identification, which captures the issues that government officials want to address through public opinion monitoring. Data were then collected from various channels, such as the "12345" government service hotline, government Weibo accounts, and official email inboxes. The risk identification phase involves the construction of a public opinion risk identification index system to identify potential risks in the data collected. This is followed by a risk assessment, where the weight of each indicator is calculated, and the risks are classified into different levels. Finally, decision recommendations were provided based on the risks identified and their urgency. The model was validated using government-citizen interaction data from Suzhou as a case study. The results of the analysis were closely aligned with the future priorities of the Suzhou municipal government, fully demonstrating the model's effectiveness and reliability of the model for early risk warning. [Results/Conclusions] The study concludes with the validation of a feasible and practical early warning model for public opinion risks. The model was tested using interaction data from the Suzhou municipal government's official website, demonstrating its effectiveness in identifying and predicting public opinion risks. The results show that the model can accurately assess the severity of risks and provide timely warnings, helping government decision-makers to manage risks proactively.
public-political interaction / risk warning / evidence-based policy making {{custom_keyword}} /
Table 2 Indicator indices表2 指标指数 |
指标 | 诉求类型 | 文本长度 | 敏感度 | 情感倾向 | 聚集度 |
---|---|---|---|---|---|
诉求类型 | 1 | 1 | 0.143 | 0.2 | 0.167 |
文本长度 | 1 | 1 | 0.143 | 0.2 | 0.167 |
敏感度 | 7 | 7 | 1.000 | 2.0 | 1.111 |
情感倾向 | 5 | 5 | 0.500 | 1.0 | 0.500 |
聚集度 | 6 | 6 | 0.900 | 2.0 | 1.000 |
Table 3 Results of AHP hierarchical analysis表3 AHP层次分析结果 |
项 | 特征向量 | 权重值/% | 最大特征根 | CI值 |
---|---|---|---|---|
诉求类型 | 0.247 | 4.941 | 5.027 | 0.007 |
文本长度 | 0.247 | 4.941 | ||
敏感度 | 1.849 | 36.989 | ||
情感倾向 | 1.204 | 24.075 | ||
聚集度 | 1.453 | 29.054 |
Table 4 Public opinion risk scale表4 舆情风险等级表 |
一级 (绿色) | 二级 (黄色) | 三级 (红色) |
---|---|---|
一般风险 | 中度风险 | 高度风险 |
Table 5 Public opinion risk factor indicator categories and quantification表5 舆情风险因素指标类别及量化方式 |
指标 | 类别代码 | 量化方式 |
---|---|---|
诉求目的 | 热线(1)、咨询(2)、建议(3)、投诉(4) | 按照公众监督栏目中填写者自己选择的诉求目的来划分 |
敏感度 | 无敏感(1)、低敏感(2)、中度敏感(3)、高敏感(4) | 在现有的评论敏感词库基础上进一步添加具有政民互动特征的敏感词(如上访等),将诉求的主要内容与新构建的敏感词库相比对,不包含任何敏感词的视为无敏感,包含1个到3个敏感词的视为低敏感,包含4个到8个敏感词的视为中度敏感,包含9个及以上的视为高度敏感 |
聚集度 | 低(1)、中(2)、高(3) | 利用余弦相似度算法判断公众诉求是否反映的是同一个问题,一个问题被仅被提到1次标记为低,被提到2至3次标记为中,被提到4次及以上标记为高 |
情感倾向 | 正向(1)、中性(2)、负向(3) | 依据SnowNLP对诉求的主要内容计算情感分数,该分数表示了文本的情感急性,低于0.3为负向情感,0.3~0.7为中性情感,高于0.7为正向情感 |
文本长度 | 短篇(1)、中篇(2)、长篇(3) | 统计公众诉求主要内容的字数,将所有诉求的字数统计结果按照升序排序,按照3:4:3的比例将文本长度划分为短篇、中篇、长篇3类 |
Table 6 Final weights of indicators表6 指标最终权重 |
指标名 | Ahp法 | Critic权重法 | 最终权重 |
---|---|---|---|
敏感度 | 0.395 92 | 0.093 489 | 0.248 350 979 |
情感倾向 | 0.205 05 | 0.238 561 | 0.328 214 030 |
文本长度 | 0.071 12 | 0.284 520 | 0.135 769 578 |
聚集度 | 0.237 57 | 0.151 238 | 0.241 074 039 |
诉求目的 | 0.089 74 | 0.232 192 | 0.139 807 748 |
Table 7 Risk level for each topic表7 各主题风险级别 |
级别 | 一级 (绿色) | 二级 (黄色) | 三级 (红色) |
---|---|---|---|
分数 | 0.2以下 | 0.2~0.5 | 0.5以上 |
包含主题 | 科教文体 交通出行 | 劳动人事 民生服务 医疗卫生 住房保障 | 环境保护 民政社区 |
Fig.3 Risk distribution of public opinion on environmental protection topics图3 环境保护主题舆情风险分布 |
1 |
中国政府网. 国务院关于加强数字政府建设的指导意见[EB/OL]. [2024-03-15].
{{custom_citation.content}}
{{custom_citation.annotation}}
|
2 |
张莉曼, 吴鹏, 尹熙成, 等. 政民互动数据中公众诉求的故事化描述: 集成、重构与叙事[J]. 情报理论与实践, 2023, 46(4): 141-149.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
3 |
马亮, 郑跃平, 张采薇. 政务热线大数据赋能城市治理创新: 价值、现状与问题[J]. 图书情报知识, 2021, 38(2): 4-12, 24.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
4 |
张莉曼, 张向先, 孙绍丹. 分布式认知视角下政民互动数据的交互叙事研究[J]. 图书情报工作, 2023, 67(9): 53-62.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
5 |
周霞, 王萍, 陈为东, 等. 政府开放数据用户认知图式联结模型——数据故事视角[J]. 情报资料工作, 2021, 42(4): 64-71.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
6 |
曹艳辉. “适度压力型”政民互动: 基于中部省级网络问政平台的数据分析[J]. 新闻与传播评论, 2023, 76(2): 70-81.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
7 |
贾改革. 政民互动中社会诉求主题挖掘和情感分析——基于人民网领导留言板的数据分析[D]. 杭州: 浙江大学, 2023.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
8 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
9 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
10 |
孙倬, 赵红, 王宗水. 网络舆情研究进展及其主题关联关系路径分析[J]. 图书情报工作, 2021, 65(7): 143-154.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
11 |
杨柳, 罗文倩, 邓春林, 等. 基于灰色关联分析的舆情分级与预警模型研究[J]. 情报科学, 2020, 38(8): 28-34.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
12 |
武慧娟, 张海涛, 王尽晖, 等. 基于熵权法的网络舆情预警模糊综合评价模型研究[J]. 情报科学, 2018, 36(7): 58-61.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
13 |
彭程, 祁凯, 黎冰雪. 基于SIR-EGM模型的复杂网络舆情传播与预警机制研究[J]. 情报科学, 2020, 38(3): 145-153.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
14 |
李知谕, 杨柳, 邓春林. 基于弹幕与评论情感倾向的食品安全舆情预警研究[J]. 科技情报研究, 2022, 4(3): 33-45.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
15 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
16 |
刘瑞, 马海群. 基于循证决策的开放数据政策制定体系构建[J]. 现代情报, 2020, 40(8): 129-133.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
17 |
马小亮, 樊春良. 基于证据的政策: 思想起源、发展和启示[J]. 科学学研究, 2015, 33(3): 353-362.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
18 |
NAM T,
{{custom_citation.content}}
{{custom_citation.annotation}}
|
19 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
20 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
21 |
魏景容. 大数据时代循证决策研究: 一个分析框架[J]. 中国科技论坛, 2020(7): 24-32.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
22 |
叶艳, 吴鹏. 循证决策视角下的患者健康咨询主题分析[J]. 情报理论与实践, 2022, 45(2): 198-203, 190.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
23 |
何玉仙. 大数据时代政府循证决策模式探究[J]. 信息系统工程, 2023(8): 120-123.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
24 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
25 |
范逢春. 国家治理现代化场域中的社会治理话语体系重构——基于话语分析的基本框架[J]. 行政论坛, 2018, 24(6): 109-115.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
26 |
王瑶. 网络问政平台中公众诉求表达对政府回应的影响研究[D]. 成都: 电子科技大学, 2024.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
27 |
赵国洪, 刘伟章. 我国政府网站政民互动模型及实证分析[J]. 情报杂志, 2012, 31(2): 195-202.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
28 |
詹承豫. 中国城市风险沟通决策的影响因素研究[J]. 治理研究, 2019, 35(5): 13-21.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
29 |
邱文. 公众诉求事件关键数据的空间智能提取与分析[J]. 城市勘测, 2020(2): 27-30.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
30 |
张楠迪扬, 郑旭扬, 赵乾翔. 政府回应性: 作为日常治理的“全回应”模式——基于LDA主题建模的地方政务服务“接诉即办”实证分析[J]. 中国行政管理, 2023(3): 68-78.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
31 |
王东琪. 网络问政的舆情挖掘及引导研究——以河北省“领导留言板”为例[D]. 石家庄: 河北经贸大学, 2023.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
32 |
梅潇, 查先进, 严亚兰. 智能推荐环境下移动社交媒体用户隐私风险感知影响机理研究[J]. 情报理论与实践, 2024, 47(1): 57-64.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
33 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
34 |
曾子明, 孙守强, 李青青. 基于融合策略的突发公共卫生事件网络舆情多模态负面情感识别[J]. 情报学报, 2023, 42(5): 611-622.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
35 |
臧雷振, 王栋, 仉佳璐. 社会科学研究中的敏感议题: 特征判断与应对方法[J]. 学习与探索, 2023(4): 36-42, 186.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
36 |
王磊, 高茂庭. 基于CRITIC权与灰色关联的隐写分析算法综合评估[J]. 计算机工程, 2017, 43(4): 154-159.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
37 |
夏立新, 杨元, 周鼎. “双一流”建设视阈下我国高校文献资源保障水平评价指标体系构建研究[J]. 图书情报工作, 2022, 66(7): 57-65.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
38 |
苏州市行政审批局. “苏州12345”2023年度工作情况发布[EB/OL]. [2024-05-12].
{{custom_citation.content}}
{{custom_citation.annotation}}
|
39 |
苏州市人民政府. 2024年政府工作报告[EB/OL]. [2024-05-12].
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_ref.label}} |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
/
〈 |
|
〉 |