摘要
【目的/意义】网络社交平台已经成为重要的大众舆论场,热点信息传播的速度快,范围广。如何在信息发布早期对其热度作出精准预测,在大数据时代具有重要研究价值,对政府妥善应对舆情事件具有实践意义。【方法/过程】本文基于微博提出一种舆情热度等级评价体系,以“郑州未经沟通将870人运抵徐州”为案例,运用多种神经网络进行预测实验,充分验证该指标体系的可行性。在此基础上进行消融实验探究何种指标对舆情热度预测的影响最大。【结果/结论】实验表明,多种模型根据该指标体系所进行的训练拟合状态良好,其中以Bi-GRU模型的正确率最高,可达92.41%。消融实验显示,内容的热度和用户历史发布内容热度之间具有强相关性。实验结果能为相关部门处理舆情事件提供参考。【创新/局限】网络舆情热度分级的理论基础待完善,如何综合多媒体信息提取特征进行热度预测是未来研究可以改进的方向。
【Purpose/significance】 Social network platforms have become an important place for public opinion,and hot information spreads quickly and widely.How to predict the popularity of information in the early stage of information release has important research value in the era of big data,and has practical significance for the government to properly respond to public opinion events.【Method/process】 Based on Weibo,this paper proposes a rating system for public opinion popularity,and takes "Zhengzhou transported 870 people to Xuzhou without communication" as an example,uses a variety of neural networks to conduct prediction experiments,and fully verifies the index system.feasibility.On this basis,the ablation experiment is carried out to explore which index has the greatest impact on the prediction of public opinion heat.【Result/conclusion】 Experiments show that the training and fitting state of various models based on this index system is good,among which the correct rate of Bi-GRU model is the highest,up to 92.41%.It shows that the index system proposed in this paper is reasonable and can provide a reference for the follow-up prediction of the popularity of public opinion.【Innovation/limitation】 The theoretical basis for rating the popularity of online public opinion needs to be improved,and how to predict the popularity by integrating the features of multimedia information extraction is a direction that can be improved in future research.
作者
刘润东
张鹏
秦瑞青
张霁阳
LIU Rundong;ZHANG Peng;QIN Ruiqing;ZHANG Jiyang(Research Center for Network Public Opinion Governance,China People's Police University,Langfang 065000,China)
出处
《情报科学》
CSSCI
北大核心
2024年第2期148-154,共7页
Information Science
基金
国家社会科学基金项目“基于大数据的网络舆情全息建模与决策情报感知研究”(20BXW074)。
关键词
网络舆情
热度预测
神经网络
深度学习
早期预测
internet public opinion
popularity prediction
neural network
deep learning
early prediction
