基于BSP-HAR-RV模型的最优抽样频率研究

Optimal Sampling Frequency based on BSP-HAR-RV Model

金融高频时间序列数量大、周期短、信息丰富,可以很好地反映金融市场特征。通过绘出平均双幂变差已实现波动率散点图(Bi Powe realized volatility Signature Plot,BSP),建立BSP-HAR-RV模型,改进以往国内通过列举法选择最优频率的方法。最后对TCL集团股票价格的高频数据进行实证分析,验证模型结果 ,并将其在最优频率下得到的HAR-RV模型预测结果与以往广泛使用的5min、10min频率得到的结果进行比较,发现最优抽样频率下模型预测能力较好,具有可行性。

High-frequency time series can reflect the characteristics of financial markets more comprehensively because of large quantities , short cycles and abundant information. By drawing an average BiPower realized volatility Signature Plot （BSP） and setting up BSP - HAR - RV model, domestic optimal-frequency-selecting method is improved. Then an empirical analysis with the TCL stock price high-frequency data is done to verify the model. Furthermore, the predict ability of the BSP-HAR-RV model based on the optimal sampling frequency is better compared to that of the HAR-RV model based on 5-min- frequency and 10-min- frequency.