| 王柯萱,高烨,李松.基于情感分类和库普曼算子的网络舆情热度预测[J].数字图书馆论坛,2025,21(8):11~22 |
| 基于情感分类和库普曼算子的网络舆情热度预测 |
| Heat Prediction of Online Public Opinion Based on Sentiment Classification and Koopman Operator |
| 投稿时间：2025-06-27 |
| DOI：10.3772/j.issn.1673-2286.2025.08.002 |
| 中文关键词: ：网络舆情；热度预测；库普曼算子；情感分类；情感因素 |
| 英文关键词: Online Public Opinion; Heat Prediction; Koopman Operator; Sentiment Classification; Sentiment Factor |
| 基金项目: |
| 作者 | 单位 | | 王柯萱 | 河北大学管理学院 | | 高烨 | 河北大学管理学院 | | 李松 | 河北大学管理学院 |
|
| 摘要点击次数: 30 |
| 全文下载次数: 41 |
| 中文摘要: |
| 为充分挖掘情感因素在舆情监测和管控中的作用，精准预测网络舆情热度趋势，本研究提出一种基于情感分类和库普曼算子的网络舆情热度预测方法。首先，构建基于BERT-CGA多特征融合的情感分类模型，充分提取情感特征并学习特征间关联，对博文内容进行情感极性分类；其次，将情感特征和库普曼算子理论引入舆情热度预测，构建融合情感因素的LSTM-EDMD热度预测模型，以LSTM算法为观测函数实现低维舆情热度指标数据到高维空间的映射，通过EDMD算法生成库普曼算子刻画系统动态特性实现热度预测，增强模型可解释性；最后，以银川烧烤店爆炸事件为例进行实证研究，实验结果验证情感因素在网络舆情热度预测中的关键作用，同时表明所提方法具有较高的预测准确性。本研究有助于政府、企业等舆情应对主体及时捕捉情感动向，准确预测舆情热度演变并制定针对性策略。 |
| 英文摘要: |
| To fully explore the role of sentiment factors in public opinion monitoring and control, and accurately predict the trend of online public opinion heat, a prediction method for online public opinion heat based on sentiment classification and the Koopman operator is proposed. Firstly, a sentiment classification model based on BERT-CGA multi-feature fusion is constructed to fully extract features and learn the correlation between features, and classify the sentiment polarity of the Weibo content. Secondly, the sentiment characteristics and Koopman operator theory are introduced into the prediction of public opinion heat. A heat prediction model of LSTM-EDMD integrating sentiment factors is constructed. The LSTM algorithm is used as the observation function to realize the mapping of low-dimensional public opinion heat index data to the high-dimensional space. The Koopman operator is generated through the EDMD algorithm to describe the dynamic characteristics of the system for heat prediction, and to enhance the interpretability of the model. Finally, an empirical study is conducted using the explosion incident at a barbecue restaurant in Yinchuan as an example. The experimental results verify the crucial role of sentiment features in predicting the heat of online public opinion and demonstrate that the prediction method proposed in this paper has high prediction accuracy. This research can assist public opinion response entities such as governments and enterprises in promptly capturing sentiment trends, accurately predicting the evolution of public opinion heat, and formulating targeted strategies. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
