大维数据背景下金融协方差阵的估计及应用

发布时间：2018-02-01 02:47

本文关键词： 乔列斯基分解法惩罚函数非参数DCC模型大维协方差阵　出处：《系统工程理论与实践》2017年03期 　论文类型：期刊论文

【摘要】：协方差阵在投资组合和风险管理中扮演着重要角色,但是大维数据给传统的协方差阵估计方法带来了巨大挑战.本文将改进的乔列斯基分解和惩罚函数等非参数方法应用到DCC模型的估计中,提出了非参数DCC模型(NPDCC).NPDCC模型首先通过改进的乔列斯基分解方法,将DCC模型估计中复杂的协方差阵估计问题转化为一系列的回归模型,然后通过引入惩罚函数,将一些回归系数压缩为零,解决了维数诅咒问题,使得大维动态条件协方差阵的估计成为可能.通过模拟和实证研究发现:较DCC模型而言,NPDCC模型明显提高了大维协方差阵的估计和预测效率;并且将其应用在投资组合时,投资者获得了更高的投资收益和经济福利.
[Abstract]:Covariance matrix plays an important role in portfolio and risk management. However, the large dimension data brings a great challenge to the traditional covariance matrix estimation method. In this paper, the improved nonparametric methods such as Cholesky decomposition and penalty function are applied to the estimation of DCC model. A nonparametric DCC model named NPDCCU. NPDCC model is proposed. Firstly, the modified Cholesky decomposition method is proposed. The complex covariance matrix estimation problem in DCC model estimation is transformed into a series of regression models, and then some regression coefficients are reduced to zero by introducing penalty function, which solves the dimension curse problem. It is possible to estimate the large dimensional dynamic conditional covariance matrix. Through simulation and empirical research, it is found that compared with the DCC model, the estimation and prediction efficiency of the large dimensional covariance matrix is improved obviously. And when applied to the portfolio, investors gain higher investment returns and economic benefits.
【作者单位】：贵州财经大学数学与统计学院;
【基金】：贵州省教育厅2015年度普通本科高校自然科学研究项目(黔教合KY字[2015]423) 国家社会科学基金(16CTJ013)~~
【分类号】：F224.0;F830.91
【正文快照】： i引言 21世纪是信息爆炸的时代,计算机技术的飞速发展,极大方便了数据的获取和存储,许多科学研究领域产生了多种多样的复杂超高维海量大数据.在金融领域中,(超)高维金融资产组合的构建对于很多个体和金融机构来说非常常见.在投资组合中,各个资产权数的确定是与它们之间的协方

【相似文献】