融入舆情发展阶段特点的热度预测  

Hotness prediction of public opinion integrated with characteristics of development stage

作　　者：解俊 马常霞 仲兆满 赵雪峰 胡文彬 XIE Jun;MA Chang-xia;ZHONG Zhao-man;ZHAO Xue-feng;HU Wen-bin(School of Computer Engineering,Jiangsu Ocean University,Lianyungang 222005,China)

机构地区：[1]江苏海洋大学计算机工程学院,江苏连云港222005

出　　处：《计算机工程与设计》2025年第3期826-833,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(72174079)。

摘　　要：为高效计算舆情评价指标权重,并考虑融入舆情发展阶段特点信息对于舆情热度预测的影响,利用高斯混合模型定量计算各指标的权重,求得较精准的舆情热度值,在此基础上,结合舆情发展各阶段特点信息,建立基于CNN-LSTM的多变量舆情热度预测模型CNN-LSTM-STAGE(CLS)。以新浪微博为平台,选取“扬州疫情”等4个网络热门舆情事件进行实例分析,预测舆情热度趋势。实验结果表明,融入舆情发展阶段特点信息能有效预测舆情的热度趋势,为舆情管控提供决策支持。To efficiently calculate the weight of public opinion evaluation index,and considering the influence of the characteristic information in the development stage on public opinion hotness prediction,a Gaussian mixture model was proposed to quantitatively calculate the weight of index to obtain a more accurate value of public opinion hotness.On the basis,combined with the characteristic information of each stage of public opinion development,a multivariate public opinion hotness prediction model CNN-LSTM-STAGE(CLS) based on CNN-LSTM was established.Taking Sina Weibo as a platform,four popular online public opinion events such as Yangzhou epidemic were selected for case analysis to predict the trend of public opinion hotness.Experimental results show that the characteristic information integrated into the development stage of public opinion can effectively predict the hotness trend of public opinion,providing the decision support for public opinion control.

关 键 词：舆情热度预测 多变量时间序列预测 高斯混合算法 卷积神经网络 长短时记忆网络 舆情阶段信息 融合模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]