摘要
从大数据视角构建高效的投资者情绪指数是行为金融投资研究的热点。本文运用文本大数据获取与情感分析技术构建股票投资者情绪指数,并从金融计量与股价预测角度检验其有效性。首先,利用网络爬虫技术抓取东方财富网的上证指数股吧实时帖文大数据,并基于文本语义情感分析构建股票投资者情绪指数;其次,对该情绪指数进行关于投资者情绪经典代理变量的回归分析,从金融计量视角分析该指数作为投资者情绪测度指标的有效性;然后,按照是否包含情绪指数因子的对比预测方式分别进行建模,利用CEEMDAN-LSTM方法对上证指数序列进行“分解—重组—预测”,进而结合预测结果深入分析不同时间尺度下情绪指数对上证指数的预测能力,从股价预测视角检验情绪指数的有效性。最后,从金融计量与股市指数预测双重视角实证检验投资者情绪指数的有效性,并证实了在不同时间尺度下该指数均可有效反映投资者对股市发展趋势预期的情绪信息。
The construction of an efficient investor sentiment index from the perspective of big data is a hot topic in behavioral financial investment research.A new stock investor sentiment index is constructed by using web crawler and text sentiment analysis,and its effectiveness on the stock market is deeply studied from both perspectives of financial measurement and stock price prediction.Firstly,the real-time posting text data of Shanghai stock index bar is accessed by web crawler,and the investor sentiment index is constructed through text semantic analysis.Then,the regression analysis of the sentiment index on classical investor sentiment proxy variables is conducted,and its effectiveness as an investor sentiment indicator is analyzed by financial measurement.Finally,comparing prediction models with and without the sentiment index factor are built,in which CEEMDAN-LSTM are used to decompose,restructure and predict Shanghai stock index,and the prediction ability of the sentiment index for Shanghai stock index under different time scales is deeply analyzed by comparing prediction effects.The empirical results demonstrate that our sentiment index can effectively reflect investors’expected information on stock market trend under different time scales,from both perspectives of financial measurement and stock price prediction.
出处
《价格理论与实践》
北大核心
2022年第11期146-151,共6页
Price:Theory & Practice
基金
教育部人文社会科学研究青年基金项目“大数据深度学习预测建模的股票在线算法交易策略研究”(21YJCZH030)
江苏省高校哲学社会科学研究项目“基于时间序列数据挖掘的股票算法交易”(2020SJA1707)
作者简介
通讯作者:贺毅岳
