摘要
石油在全球经济发展中扮演着重要的角色.为了更有效地预测石油价格,本文将新闻媒体和舆论事件等信息纳入到油价预测研究中.首先,应用主题识别模型对与油价相关的网络新闻进行主题识别,将其转化为主题分布数据.并根据新闻主题对新闻进行情感分析,最后将主题分布与每个主题下的新闻情感特征加入到油价预测模型中,以此来改进预测效果.实证结果表明,相较于基准模型,结合了网络新闻主题特征以及主题分类下新闻情感特征的预测模型具有更好的预测性能.这表明,新闻媒体中蕴含的信息能够有效反映市场情绪,并有助于对油价进行预测.
Crude oil plays an important role in global economic development.To effectively predict crude oil prices,this paper incorporates news media into oil price forecasting.First,we apply a topic recognition model to identify the topics of online news and transform them into topic distribution data.Then we conduct sentiment analysis on news articles based on their topics,finally,we integrate the topic distribution and news sentiment features under each topic into the oil price prediction model to enhance its accuracy.The empirical results indicate that integrate network news topic features and news sentiment features under topic classification performs better than the benchmark model.This suggests that the information contained in news media can effectively reflect market sentiment and contribute to the forecasting of crude oil prices.
作者
余乐安
赵晨珊
宋正阳
YU Lean;ZHAO Chenshan;SONG Zhengyang(Business School,Sichuan University,Chengdu 610065,China;School of Economics and Management,Beijing University of Chemical Technology,Beijing 100029,China)
出处
《计量经济学报》
CSSCI
CSCD
2023年第2期443-463,共21页
China Journal of Econometrics
基金
国家自然科学基金重点项目(72034003)。
关键词
网络新闻
主题分布
情感分析
文本挖掘
油价预测
online news
topic distribution
sentiment analysis
text mining
crude oil price forecasting
