危机情境下网络反向议程设置的探测及预测  

Detection and Prediction of Online Reverse Agenda-setting in Crisis Situation

作　　者：张丽 安璐[1] 阮雪琴 Zhang Li;An Lu;Ruan Xueqin(School of Information Management,Wuhan University,Hubei Wuhan 430072)

机构地区：[1]武汉大学信息管理学院,湖北武汉430072

出　　处：《情报理论与实践》2024年第8期129-139,119,共12页Information Studies:Theory & Application

基　　金：国家社会科学基金重大项目“不确定环境下韧性社会智能情报支持与决策研究”的成果,项目编号:23&ZD230。

摘　　要：[目的/意义]反向议程设置探测方法能够动态捕捉危机情境下局部而短暂的反向议程设置效应,反向议程设置的预测模型有助于理解危机信息从公众议程进入媒体议程的过程,能够有效追踪舆情动态并辅助危机报道的选题。[方法/过程]采用滑动时间窗口和格兰杰因果检验,动态探测危机情境下的反向议程设置效应;基于新闻价值和资源动员理论,从内容价值、媒体差异、话语风格、动员举措和成员网络5个方面提出反向议程设置的特征体系,采用逻辑回归、朴素贝叶斯、支持向量机、CART决策树和XGBoost训练反向议程设置预测模型,使用XGBoost解释特征重要性排序,利用SHAP解释特征影响。[结果/结论]公众议程与媒体议程的领先关系随着危机议题的发展动态变化,媒体议程影响公众议程的速度高于公众议程影响媒体议程的速度。危机情境下基于XGBoost的反向议程设置预测模型效果最好,准确率达到91.04%,F1值达到90.80%,AUC值达到96.58%;内容价值、媒体差异特征对预测结果重要性最高。[Purpose/significance]The reverse agenda-setting detection method can dynamically capture the partial and transient reverse agenda-setting effects in crisis situations.The prediction model of reverse agenda setting helps to understand the process of crisis information entering the media agenda from the public agenda,and can effectively track public opinion dynamics and assist in selecting topics for crisis reporting.[Method/process]In this study,we employed sliding time windows and Granger causality test to dynamically detect the reverse agenda-setting effect in crisis situations.Drawing on news value and resource mobilization theory,we proposed a comprehensive feature system for reverse agenda-setting,encompassing content value,media differentiation,discourse style,mobilization measures,and member network.The reverse agenda-setting prediction model was trained using logistic regression,naive Bayes,support vector machine,CART decision tree,and XGBoost algorithms.XGBoost was utilized to elucidate feature importance ranking,while SHAP was employed to explicate feature influence.[Result/conclusion]The leading relationship between public agenda and media agenda changes dynamically with the development of crisis issues,and the speed at which media agenda affects public agenda is higher than the speed at which public agenda affects media agenda.In the crisis situation,the reverse agenda-setting prediction model based on XGBoost has the best effect,with an accuracy of 91.04%,an F1 value of 90.80%,and an AUC value of 96.58%.Content value and media differentiation features are the most important for predicting reverse agenda-setting.

关 键 词：反向议程设置 危机情境 注意力分配 格兰杰因果分析 XGBoost 

分 类 号：G206[文化科学—传播学]