The Estimation Of High-frequency Data Volatility Based On Local Mean Decomposition

With the development of science and technology,analysis of financial high-frequency data has quickly become a hotspot and difficulty in the field of data analysis.High-frequency data contains more market detailed information than low-frequency data,so the volatility estimate for high-frequency data must be a more realistic description of market fluctuations.As one of the most important features of financial markets,volatility can accurately depict the volatility of returns,measure the risks in the financial market,and judge the level of returns.The realized volatility has been studied by more and more scholars due to its features of no model and convenient calculation,and the related methods of improvement and expansion have also emerged in an endless stream.Due to the existence of microstructure noise in financial high frequency data,it's need to be processed in advance and then estimate the volatility.This paper proposes a method based on local mean decomposition to estimate the volatility.The empirical analysis uses the 1-min trade data for the same quarter of five stocks.This paper has two parts.The first part is denoising research.First,simulation data is used to verify the denoising effect of local mean decomposition and empirical mode decomposition respectively.Then the better denoising method is used to denoise the empirical data.The second part is the estimation of volatility.Firstly,using simulation data to explore the performance of local mean decomposition method to estimate volatility,then a preliminary inference was obtained,i.e.the higher the data frequency is,the lower the estimation accuracy is.Second,use the local mean decomposition and Hilbert-huang transform to estimate the volatility of the denoised 1-min data.In addition,the data was sampled obtain 5-min,15-min,and 30-min data.The above two different methods were also used to estimate the volatility of these different time interval data.Finally,compare the results of estimated volatility using the two methods in different time interval and observe the relative error between the two to determine which method has higher estimation accuracy.The empirical results show that the estimated accuracy of the local mean decomposition method is higher than the Hilbert-Huang transform method for the 1-min,5-min,and 15-min data.However,for the estimation of 30-mine data,the local mean decomposition method has lower estimation accuracy than the Hilbert-huang transform.In general,the relative error of the LMD method increases with the increase of the time interval,which is consistent with the results of the simulation.