摘要
有效提取股票价格时间序列中股票对之间的相互依赖关系能够提高投资组合的收益率。采用基于块坐标下降法的非负张量分解技术从股票价格时间序列中提取复杂关系,构建预测距离矩阵来代替原有的相关系数矩阵,提出基于非负张量分解的投资组合策略。选取2019—2021年中证100指数数据进行实证分析,实验结果表明:基于非负张量分解的投资组合策略具有较高的可行性,且在股市动荡时期表现要优于等权重和市值加权投资组合。
Effective extraction of the inter dependence between the stock pairs from stock price time series can improve the return rate of portfolio investment.This study uses non-negative tensor decomposition technology based on block coordinate descent method to extract complex relationships from stock price time series,constructs prediction distance matrix to replace the original correlation coefficient matrix and proposes a portfolio strategy based on non-negative tensor decomposition.Through selecting the China securities index(CSI)100 Index dataset from 2019 to 2021 for empirical analysis,the experiment shows that the portfolio strategy based on non-negative tensor decomposition is feasible and performs better than equal-weight and market-cap-weighted portfolios in times of stock market turmoil.
作者
徐相建
马海洋
赵为华
XU Xiangjian;MA Haiyang;ZHAO Weihua(School of Sciences,Nantong University,Nantong 226019,China)
出处
《南通大学学报（自然科学版）》
CAS
2023年第2期79-85,共7页
Journal of Nantong University(Natural Science Edition)
基金
国家自然科学基金面上项目(11971171)
国家社会科学基金项目(22BTJ025)。
关键词
非负张量分解
投资组合
块坐标下降法
nonnegative tensor decomposition
portfolio strategy
block coordinate descent method
