摘要
为了利用因子排序组合的信息并保证组合权重具有一定的稀疏性,基于Sparse Group Lasso (SGLasso)和经典的均值-方差(mean-variance,MV)投资组合策略,构建了能够对高维资产数据集进行投资的SGLasso-MV策略.与Lasso和GLasso相比,SGLasso能够同时实现组内和组间的稀疏性,并利用了特征分组信息,因此适用于改进MV策略输出权重的不稳定性和高误差性问题.在实证数据方面,利用A股1997年至2019年所有可用A股股票的日际实证数据集,进行了不固定成分股的滚动投资,以避免样本选择性偏误,并将SGLasso-MV与几种经典的投资组合策略进行了比较.结果显示,相比其他同样包含期望收益率估计量的策略,SGLasso-MV的权重能够在样本外实现显著更低的标准差风险和更低的换手率.
To utilize factor characteristics sorting information and to ensure some sparsity on portfolio weight,this paper is based on Sparse Group Lasso(SGLasso) and classic mean-variance(MV) portfolio strategy,and constructs a new strategy called SGLasso-MV,which can be used to handle high-dimensional asset datasets.Compared to Lasso and GLasso,SGLasso can achieve both within-group and between-group sparsity and can use feature grouping data,therefore,it is suitable for improving the unstability and high-error issues of MV weight output.In terms of empirical data,this paper uses all available daily data on Chinese A share market from 1997 to 2019,and conducts constituent-flexible rolling investments,to avoid sample selection bias,and compares SGLasso-MV to several classic portfolio strategies.The results show that SGLasso-MV weights can achieve significantly lower out-of-sample standard error risk and turnover compared to other strategies that also require estimation of expected return.
作者
陈胜
赵慧敏
CHEN Sheng;ZHAO Hui-min(Department of Government Policy and Public Administration,University of Chinese Academy of Social Sciences,Beijing 102488,China;School of Business,Sun Yat-sen University,Guangzhou 510275,China)
出处
《数学的实践与认识》
2021年第12期323-328,共6页
Mathematics in Practice and Theory
基金
国家自然科学基金重大项目(71991474)
中央高校基本科研业务费专项资金资助(31620527)。