基于LSTM网络和文本情感分析的股票指数预测  

Stock Index Prediction Based on LSTM Network and Text Sentiment Analysis

作　　者：于孝建[1,2] 刘国鹏 刘建林[1] 肖炜麟[3] Yu Xiaojian;Liu Guopeng;Liu Jianlin;Xiao Weilin(School of Economics and Finance,South China University of Technology,Guangzhou 510006,China;Research Center of Financial Engineering,South China University of Technology,Guangzhou 510006,China;School of Management,Zhejiang University,Hangzhou 310058,China)

机构地区：[1]华南理工大学经济与金融学院,广东广州510006 [2]华南理工大学金融工程研究中心,广东广州510006 [3]浙江大学管理学院,浙江杭州310058

出　　处：《中国管理科学》2024年第8期25-35,共11页Chinese Journal of Management Science

基　　金：长三角科技创新共同体联合攻关计划项目(2022CSJGG0800);国家社会科学基金重点项目(22AZD039);中央高校基本科研业务费专项资金(QNZD202211);教育部人文社会科学研究一般项目(23YJA630102)。

摘　　要：投资决策受投资者行为偏好的影响,因此合理地捕捉投资者情绪有助于预测股票市场未来变化趋势。结合机器学习算法,分析金融市场投资者情绪,利用SVM情感分类算法,对股吧个股评论中的文本数据进行分析,从而构建出反映投资者情绪的市场情绪指标。进一步使用LSTM深度学习网络,提取市场情绪指标特征,对上证50指数进行短期预测,并对比多种传统时间序列分析模型和机器学习模型。研究结果表明,LSTM神经网络在金融时间序列预测上具有更高的准确率和精确度;加入市场情绪特征后,能进一步提升LSTM模型预测结果的准确率和精确度,说明了投资者市场情绪对于市场指数预测的有效性和适用性;此外,对LSTM模型预测结果进行误差修正,能够有效优化LSTM模型的预测结果。Investment decision-making can be a complex process,influenced by various factors,including investor behavior preferences.Therefore,it's important to understand and capture investor sentiment for predicting future changes in the stock market trend.In this regard,machine learning algorithms can be helpful in analyzing investor sentiment in the financial market.It aims to construct a predictive model for stock indices using an LSTM network and text sentiment analysis in this paper.To begin with,a web crawler program is used to collect text comments on individual stocks in the East Money Stock Bar.The text data are analyzed using the SVM sentiment classification algorithm to construct a market sentiment index that reflects investor sentiment.Additionally,the LSTM deep learning network is used to extract the features of the market sentiment index and make short-term predictions on the SSE 50 index.Various traditional time series analysis models and machine learning models are compared.The results show that the LSTM neural network has higher accuracy and precision in financial time series prediction.After incorporating market sentiment features,the accuracy and precision of the LSTM network prediction results can be improved.This indicates that investor market sentiment is highly effective and applicable for market index prediction.It is also found that error correction of the LSTM network prediction results can effectively optimize the prediction results.Overall,a new method is provided for understanding investor sentiment and predicting future changes in the stock market trend.It is hoped that our research results can provide useful reference and guidance for financial investors and analysts.

关 键 词：LSTM神经网络 SVM分类器 金融文本 市场情绪指标 股票指数预测 

分 类 号：F830.91[经济管理—金融学]