多维金融高频协方差阵预测模型的比较分析

现代投资组合理论大部分是从组合风险控制的角度展开,协方差矩阵扮演着非常重要的角色.将高频协方差阵应用在投资组合或风险管理时,就需要考虑采用何种预测模型来对高频协方差阵进行预测,较好的预测模型能够更加准确的对资产的波动性进行预测.高频协方差阵预测模型的建立较为复杂,目前还没有一种广泛被认可的模型.采用MCS检验法来选择最优的预测模型,研究发现高频协方差阵预测模型LOG-HAR模型在所有的损失函数下预测能力最好,并且高频协方差阵预测模型的预测能力要优于低频协方差阵预测模型.

The Comparative Analysis of Multidimensional Financial High Frequency Covariance Matrix's Prediction Model

modern portfolio theory mostly expand from the perspective of portfolio risk control,covariance matrix plays an important role.When applying high frequency covariance matrix to the portfolio and risk managing,we should consider what kind of prediction model be adopted to predict the high-frequency covariance matrix.Better prediction model can be more accurate for predicting the volatility of asset.High-frequency covariance matrix's prediction model is very complex,there is no a widely recognized model now.This paper will use the MCS test to choose the best prediction model.Through the research,we found that LOG-HAR model has the best predictive ability under all of the loss function.And the predictive abihty of the high-frequency covariance matrix forecasting model is superior the low-frequency prediction model.