一种基于时间序列链的股市预测方法  

Research on stock market prediction method based on time series chain

作　　者：田贤忠[1] 胡安娜 顾思义 TIAN Xianzhong;HU Anna;GU Siyi(College of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310023,China)

机构地区：[1]浙江工业大学计算机科学与技术学院,浙江杭州310023

出　　处：《浙江工业大学学报》2021年第5期503-510,563,共9页Journal of Zhejiang University of Technology

基　　金：国家自然科学基金资助项目(61672465)。

摘　　要：随着互联网的发展,如何在海量的数据中挖掘出有益的信息尤为重要。利用时间序列模型预测股票市场虽然早已被证实是有效性的,但是过去都停留在时间序列主题发现上,忽视了子序列也会随时间动态演化的情况。因此,基于近年提出的新概念——时间序列链,提出将其应用在股票市场预测中的算法,结合股票市场的特点,利用向量的余弦距离来模拟股票走势对子序列相似度的影响,同时提出基于CEEMDAN的子序列长度搜索算法来获取尽可能有效且有代表性的子序列长度,根据这些子序列长度来查找时间序列链并应用于股票预测。历史股票数据回测证明笔者算法的预测准确率在93%以上。With the development of the internet,how to dig out useful information from massive amounts of data is particularly important.The use of time series models to predict the stock market has long been proved to be effective,but in the past,it has been stuck on the discovery of time series topics,ignoring the situation that sub-sequences will also dynamically evolve over time.Therefore,based on a new concept that has just been proposed in recent years-time series chain,an algorithm that applies it to stock market prediction is proposed,combined with the characteristics of the stock market,and the cosine distance of the vector is used to simulate the impact of stock trends on the similarity of sub-sequences.At the same time,a CEEMDAN-based sub-sequence length search algorithm is proposed to obtain as effective and representative sub-sequence length as possible.According to these sub-sequence lengths,the time series chain is searched and applied to stock prediction.Back-testing of historical stock data proves that the prediction accuracy of the author s algorithm is above 93%.

关 键 词：股市预测 时间序列模型 时间序列链 CEEMDAN 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]