中文摘要：

采用上证综指2000-2008年的高频数据，在考察了中国股市已实现波动率的特征（即具有长记忆性、结构突变、不对称性和周内效应的特征并且结构突变只能部分解释已实现波动率的长记忆性）的基础上，构建了一个自适应的不对称性HAR－D—FIGARCH模型，并用于波动率的预测。模型的估计结果表明，与其他HAR模型相比，该模型对样本内数据的拟合效果最好。最后，通过SPA检验实证评价和比较了该模型与其他5种已实现波动率预测模型的样本外预测精度。结果发现，在各种损失函数下，该模型是预测中国股市已实现波动率精度最高的模型。

英文摘要：

We explore the characteristics of the high-frequency volatility in Chinese stock markets by employing the high-frequency volatility data from SSEC, and find that the volatility has long term memory, structural breaks, asymmetry, day-of-the-week effect. In addition, structural breaks can only partially explain the long memory. To capture these characteristics simultaneously we proposes an adaptive asym- metry HAR-D-FIGARCH model and use it to conduct a validity forecast. As compared with other HAR model, the proposed model improves the in-sample fitting significantly. As compared to other 5 models by utilizing SPA test, we find that, under various loss functions, the proposed model is the best model for high-frequency volatility forecasts among the 6 models in Chinese stock markets.