摘要
如何提升多因子模型对超额收益的经济解释同时兼顾实际的投资效果一直是资产定价和投资学研究领域的热点和棘手问题之一.在因子的提取问题上,大部分实证研究局限在改变或替换因子上.将机器学习中的K-NN最临近点算法应用到因子提取之中,不仅提升了多因子模型的经济解释能力,而且保证了较好的实证结果的统计性质,进而据此给出了一个量化选股策略.通过选取沪深300指数全部股票从2014年至2019年的相关数据进行实证分析,并使用2020年1月至9月的数据进行了外推及效果验证.实证结果表明,给出的因子提取方法具备更好的统计检验效果且能得到更高的累计收益,并且量化选股策略具有更优秀的收益风险表现.
With the maturity and enrichment of factor investment models,a large amount of domestic and overseas relevant studies among theory and demonstration have emerged.However,the emergence of these studies is companied with the problem that how to improve the interpretation of excess returns while getting a better investment performance.There’re various multi-factor models,which are the most widely used in empirical asset pricing research,has a problem of insufficient explanatory power in the empirical analysis of Chinese A-share market.What’s more,considering the problem among factor extraction,lots of empirical research is limited to change or replace factors.This article considers using the K-NN nearest neighbor algorithm among machine learning to factor extraction and proposes a new solution to the above problems,and proposes a quantitative stock selection and allocation strategy accordingly with a comparative study.After that,this article selects the relevant data which used in of all the stocks of the SH300 Index from 2014 to 2019 as the numerical example,and the data from January to September 2020 are used for extrapolation and effect verification.The empirical and verification results show that the new scheme of factor extraction presented in this paper has a better statistical test effect and can obtain a higher cumulative return effect,and has a better return risk performance under the same stock selection strategy.
作者
金元浩
房勇
卢焱
JIN Yuan-hao;FANG Yong;LU Yan(Academy of Mathematics and Systems Science,Chinese Academy of Sciences,Beijing 100190,China;School of Economics and Management,University of Chinese Academy of Science,Beijing 100190,China)
出处
《数学的实践与认识》
2021年第19期106-119,共14页
Mathematics in Practice and Theory
基金
国家自然科学基金(71631008)。