张程 周恬恬



摘 要: 股市走势预测是金融学、统计学、机器学习等多学科交叉研究的热点学术问题。股市走势并不完全由自身内在规律决定,也会受到投资者的关注度的影响。本文研究了投资者关注与上海证券综合指数之间的波动效应,提出了一种基于百度指数并结合随机森林模型的上证指数走势预测方法。基于已有的初始搜索关键词词库在百度指数网站获取相应的关键词的百度指数,通过时差相关分析法筛选出具有预测意义的关键词,将筛选后的关键词的百度指数数据和上证指数的相关属性数据作为预测模型的输入数据,通过随机森林建立预测模型并实现对上证指数的预测。对比实验证明引入百度指数后的预测模型比传统的预测模型具有更高的准确率。
关键词: 上证指数预测;百度指数;随机森林;时差相关分析法
中图分类号: TP391 文献标识码: A DOI:10.3969/j.issn.1003-6970.2020.06.012
本文著录格式:张程,周恬恬. 基于百度指数和随机森林的上证综指预测[J]. 软件,2020,41(06):5662
【Abstract】: Stock market trend forecast is a hot academic issue that has attracted many researchers and participants in multiple disciplines such as economics, statistics, and machine learning. Stock market trend is not entirely determined by its own internal laws, and it will also be affected by the investors' attention. This paper studies the ripple effect between investors' attention and the Shanghai Composite Index, and proposes a Shanghai Composite Index forecast method based on the Baidu Index and random forest. Firstly, obtain the corresponding Baidu index of the existing initial search keyword thesaurus based on the Baidu Index website, and then screen the keyword with predictive significance by the method of time difference correlation analysis. Finally, use the keywords Baidu index and the relevant attribute data of the Shanghai Composite Index as input of forecast model that based on random forest to realize Shanghai Composite Index forecasting. The experiment results show that the proposed forecasting method has higher accuracy than the traditional method.
【Key words】: Shanghai composite index forecast; Baidu index; Random forest; Time difference correlation analysis
0 引言
时间序列预测、灰色模型和机器学习是常用的股市预测方法。时间序列预测是通过分析股票价格的过去收益来预测未来股价(Michel Ballings 2015)[1]。传统的灰色模型数学虽然容易理解且计算简单。但是,其原始模型的线性特性导致无法预测本质为非线性的急剧变化的数据(Chun-I Chen a2010)[2]。Bruno Miranda Henrique使用了支持向量机(Bruno Miranda Henrique 2018)[3],Yi Zuo使用贝叶斯网络(Zuo Y 2012)[4],Leonel A.Laboissiere基于人工神经网络进行最高和最低股价预测研究(Laboissiere L A 2015)[5]。Hsu M W证实了最好的机器学习方法比最好的计量经济学方法产生更准确的预测(Hsu M W 2016)[6]。……