Study On The Asset Allocation Problem Based On High Frequency Financial Data

Asset allocation is the core content of modern finance. Since the last century the Markowitz mean variance analysis framework, to get breakthrough in the development of the asset allocation problem in multi aspects. Although many researchers in different aspects of the Markowitz mean variance analysis framework, but the Markowitz mean variance analysis is still the most classical paradigm. Measures of risk assets returns as a core variable in the analysis paradigm has become one of the main development of the follow-up studies Markowitz model. Is generally believed that the volatility of asset returns is the return on assets risk measure index, the most intuitive best practice. Study on the volatility of asset returns is the core of financial econometrics and hot research topic in the field of. High frequency financial data, the realized volatility has a solid foundation for the mathematical thought and model free, easy to estimate, can be directly modeling has become one of the most fruitful areas based on the recent.This paper realized volatility is thought to construct the realized covariance matrix as a measure of risk asset allocation problem into the return risk analysis framework based on, with overall volatility (Integrated Volatility) as the unbiased estimator of risk assets in a period of time measurement. The establishment of asset allocation problem of high frequency financial data revenues realized covariance matrix based on the analysis framework. After a brief introduction of the origin of the asset allocation problem and the basic theory, the system of modern asset allocation theory introduces Markowitz mean variance model represented; then introduced the follow-up research especially the research on the risk measure and the latest research dynamic of the model. Then this paper introduces the research status of financial high frequency data based on field and has realized covariance matrix and its modeling volatility thought. High frequency financial data has unique properties and processing because of its high frequency. This paper introduces the characteristics and processing technology of high frequency financial data. After introduces several common deviation noise reduction technology, the volatility matrix sequence using the conditional autoregressive Wishart (Conditional Autoregressive Wishart) model of the realized covariance (Realized Covariance RCOV) modeling description. Then the rate of return volatility model and matrix model combining, establish mean -CAW model and mean -CAW-HAR model. The volatility variable is the core variable distribution of assets, the realized volatility matrix based on the latest research on the asset allocation problem based on realized volatility matrix. And on the basis of this model, makes an empirical study of ultra high frequency data using the Shanghai and Shenzhen stock markets. Get the following conclusion:conclusion:the information in the stock trading frequently in the empirical analysis of the vast majority of abandoned. With the decrease of sampling frequency, the data retention ratio dropped sharply. The multidimensional case, even in the most reserved information refresh time synchronization method for 3D data only assets, about half of the abandoned; this is an empirical analysis of the actual process of asset price especially the construction attention and solve the problem of allocation of assets based on. Conclusion two:through the statistical feature matrix sequence analysis of volatility, the volatility that is a dynamic process of continuous change. The asset price volatility assumption and empirical conclusion is inconsistent, and the dynamic process of modeling volatility. To effectively capture the volatility characteristics. Provide a solid foundation for solving the objective function of asset allocation problem. Conclusion:through three different data sets, different matrix algorithm to achieve the volatility, comparative analysis of different mean volatility model. Found that the realized volatility algorithm effects on model parameters significantly. This kind of influence in different data sets, have been demonstrated in different models.