中文摘要：

现有的金融高频数据研究，并未充分考虑微观结构噪声对波动建模和预测的影响．以非参数化方法为理论框架，基于高频数据，采用适当方法分离出波动中的微观结构噪声成份，构建了新的跳跃方差和连续样本路径方差，将已实现波动分解为连续样本路径方差、跳跃方差和微观结构噪声方差．同时考虑微观结构噪声和跳跃对波动的影响，对HAR-RV-CJ模型进行改进，提出了HAR-RV-N-CJ模型和LHAR-RV-N-CJ模型．通过上证综指高频数据进行实证，结果表明新模型在模型拟合和预测方面均优于HAR-RV-CJ模型．

英文摘要：

The effect of microstructure noise on volatility modeling and forecasting is not well enough considered by the existing researches on financial high-frequency data. Based on the theoretical framework of non-parametric approach and with high-frequency data, the microstructure noise components are separated from volatility via suitable method. The new jump variance and continuous sample patlp variance are constructed and the realized volatility is divided into continuous sample path variance, jump variance and microstructure noise variance. With considering simultaneously the effects of microstructure noise and jump on volatility, the HAR-RV-N-CJ model and LHAR-RV-N-CJ model are put forward by virtue of modifying HAR-RV-CJ model. The empirical analyses indicate that the new models outperform HAR-RV-CJ mociel in model goodness of fit and forecasting by using the high-frequency data from Shanghai composite index.