基于多视角股票特征的股票预测研究  

Research on stock forecasting based on multi⁃view stock features

作　　者：李金轩 杜军平[1] 薛哲 Li Jinxuan;Du Junping;Xue Zhe(Beijing Key Laboratory of Intelligent Telecommunication Software and Multimedia,School of Computer Science,Beijing University of Posts and Telecommunications,100876,Beijing,China)

机构地区：[1]智能通信软件与多媒体北京市重点实验室,北京邮电大学计算机学院,北京100876

出　　处：《南京大学学报（自然科学版）》2021年第1期68-74,共7页Journal of Nanjing University（Natural Science）

基　　金：国家自然科学基金(61902037,61772083,61802028);广西科技重大专项(桂科AA18118054)。

摘　　要：股票价格预测是金融行业中的一个重要研究内容,能够更准确地分析股票价格走势对于投资机构至关重要.目前,关于自动化预测股票价格发展的研究工作相对较少,还有许多问题需要解决.针对传统股票预测方法中视角单一、无法充分考虑数据的各特征重要度的问题,提出一种基于多视角股票特征的股票预测方法,通过计算股票数据的Ma,Macd,Kdj,Boll特征指标,训练每个指标下的弱学习器,并进行多个弱学习器的集成学习,最终用于预测股票价格走势.使用美国股票新闻数据集进行验证.结果表明,基于多视角股票特征的股票预测方法预测得到的股票价格与实际价格之间的平均误差与均方误差分别为1.9321和0.0581,优于传统的基于单一指标的股票预测结果.Stock price forecasting is an important research content in financial industry,and it is very important for investment institutions to analyze the stock price trend more accurately.At present,there are relatively few researches on the automatic forecasting of stock price development,and many problems need to be solved.Among them,the perspective of traditional stock forecasting methods is single so that we can't fully consider the importance of each feature and the main features of the data.In this paper,we propose a stock forecasting method based on multi⁃view stock features.By calculating the Ma,Macd,Kdj,Boll feature indexes of stock data,we train the weak learners under each index,and carry out the integrated learning of multiple weak learners.Finally,the stock price trend is predicted.The stock news data set of the United States is used for verification.The experimental results show that the average error and the mean square error between the stock price and the actual price predicted by the stock forecast method based on multi⁃view stock features are 1.9321 and 0.0581,respectively.It is superior to traditional stock forecast results based on a single index.

关 键 词：股票预测 多视角 弱学习器 集成学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]