投资组合中协方差阵的估计和预测

Estimation and Prediction of Covariance Matrix in Portfolio

为了解除噪声和跳跃以及维数诅咒对协方差阵估计的影响,将修正的门限预平均已实现协方差阵(MTPCOV)与VAR-LASSO模型相结合,估计和预测高维高频数据的协方差阵,并将其应用在投资组合中。研究发现,由VAR-LASSO模型预测的MTPCOV估计量应用在投资组合中,可以使得投资者获得更高的收益,并降低投资风险。

In order to solve the influence of noise、jump and dimension curse on covariance matrix estimation,this paper combines the modified threshold pre-averaging covariance matrix（MTPCOV）with VAR-LASSO model,to estimate and predict the covariance matrix of high-dimensional and high-frequency data.The main finding is：The MTPCOV which is predicted by the VAR-LASSO model can be used in the portfolio to gain higher profits and reduce investment risk.