基于EEMD与HAR-RV模型的已实现波动率研究

高频金融数据下对资产价格波动的研究越来越受到人们的关注，而随着对数据行为解析能力的提高，噪音也会随着采样频率的提高而增加，从而导致已实现波动率的估计偏差。为了降低高频数据中噪音对波动率估计的影响，在HAR-RV模型基础上使用EEMD结合小波分析的方法提高估计的有效性。实证研究发现，仅使用EEMD进行降噪预测并不能很好地预测股票市场的实际变动趋势，而在EEMD分解后的高频部分中使用小波方法进行处理，发现降噪后构建模型的均方误差（MSE）与平均绝对误差（MAE）分别下降了93.92%及76.94%，能够满足对股票市场实际序列变动的预测要求。

Realized Volatility Research Based on EEMD and HAR-RV Model

The study of asset price volatility in high-frequency financial data has been attracting more and more attention. With the improvement of analytic ability of data behavior， noise will also increase with the increase of sampling frequency， which will lead to the estimated deviation of realized volatility. In order to reduce the impact of noise on volatility estimation in high-frequency data， this paper uses EEMD combined with wavelet analysis to improve the estimation effectiveness on the basis of HAR-RV model. The empirical study finds that only using EEMD for noise reduction prediction could not meet the actual sequence change of the stock market， and the wavelet method is used for the high-frequency part after EEMD decomposition， and it is found that the MSE and MAE of the model constructed after noise reduction fall 93.92% and 76.94%， respectively， which can meet the prediction requirements of the actual sequence change of the stock market.