股市舆情潜藏情感倾向对收益率的预测研究  

Research on prediction of stock market yield rate with potential emotion-tendency of public sentiment on stock market

作　　者：朱昶胜[1] 张翰垠 冯文芳[2] ZHU Chang-sheng;ZHANG Han-yin;FENG Wen-fang(School of Computer and Communication, Lanzhou Univ. of Tech., Lanzhou 730050, China;School of Economics and Management, Lanzhou Univ. of Tech., Lanzhou 730050, China)

机构地区：[1]兰州理工大学计算机与通信学院,甘肃兰州730050 [2]兰州理工大学经济管理学院,甘肃兰州730050

出　　处：《兰州理工大学学报》2019年第5期109-114,共6页Journal of Lanzhou University of Technology

基　　金：兰州理工大学红柳杰出人才基金项目(J201304)的资助

摘　　要：以网络股评舆情数据作为非结构型文本数据研究对象,结合股票市场的相关交易指标,使用文本挖掘技术和机器学习算法确定投资者情绪测度指标,分析舆情数据中潜藏情感倾向对未来短期内股票收益率的预测能力.实证结果表明,舆情文本中挖掘潜藏情感信息能够以较高的准确率实现对股市收益率的预测.分析讨论了对预测结果有一定影响的特征字段与训练样本两个因素,发现在特征字段数量不变的情况下,随着训练数据的增多,预测结果的解释能力会有所提高;而当训练数据维持在一定范围内时,特征词数量的选取对预测结果也有很大的影响.Taking the public sentiment data of network stock assessment as the research object of unstructured text data and combining the relevant indices of stock market, the sentiment measuring indices of investor are determined with text mining technique and machine learning algorithm and the prediction ability of future short-term stock market yield rate by means of potential emotion-tendency in public sentiment data is analyzed. It is shown by actual proof that the mining of potential emotional information in public sentiment text can predict the stock market yield with a high accuracy. The two factors-characteristic word section and training sample, which have a certain influence on the prediction result, are analyzed and discussed. It is found that the interpretation ability of the prediction results will be improved to some extent with the increase of the training data in the case of unchanged amount of characteristic word section. When the training data is maintained within a certain range, the selection of characteristic word amount will have a great influence on prediction result, also.

关 键 词：潜藏情感倾向 文本挖掘 舆情数据 机器学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]