摘要
波动矩阵在投资组合和风险管理中扮演着重要角色,但在高维数据背景下,维数诅咒和噪声的影(IMP-MFD)。该模型将波动矩阵分解为资产的标准差(波动的平方根)构成的对角矩阵与资产间相关矩阵的乘积;然后采用基于高频数据的、预测精度更高的FNN-HAR-J模型来估计和预测对角波动矩阵,采用基于低频数据的、能够很好地解决维数诅咒和噪声影响的POTDCC模型来估计和预测动态相关性矩阵。通过实证和稳健性分析得知:IMP-MFD模型明显优于MFD模型,将其应用于投资组合可以使投资者获得更高的收益。
Volatility matrix plays an important role in portfolio and risk management. However, in the context of high-dimensional data, the influence of dimensionality curse and noise make it difficult to estimate the volatility matrix. On the basis of previous studies, this paper proposes an improved mixed-frequency data model(IMP-MFD). In the proposed model, the volatility matrix is decomposed into the product of the diagonal matrix composed of the standard deviation of assets(the square root of volatility) and the correlation matrix between assets.Then, the FNN-HAR-J model with higher prediction accuracy based on high-frequency data is used to estimate and predict the diagonal volatility matrix. Finally, the low frequency data-based POTDCC model,which can solve the influence of dimensionality curse and noise, is used to estimate and predict the dynamic correlation matrix.The empirical and robust analysis shows that the IMP-MFD model is obviously superior to the MFD model, and the application of IMP-MFD model in the portfolio can enable investors to obtain higher returns.
作者
刘丽萍
Liu Liping(School of Mathematics and Statistics,Guizhou University of Finance and Economics,Guiyang 550025,China)
出处
《统计与决策》
CSSCI
北大核心
2021年第14期132-136,共5页
Statistics & Decision
基金
贵州省科技厅项目(黔教合基础[2019]1050号)
贵州省教育厅人文社会科学项目(2019zc082)
贵州省教育厅自然科学项目(黔教合KY字[2018]160号)。
关键词
混频数据
高维波动矩阵
IMP-MFD模型
mixed-frequency data
high-dimensional volatility matrix
IMP-MFD model
