Modeling Of Financial Market's (Ultra) High-Frequency Data And Comparative Study With Low-Frequency Data

In the field of financial market microstructure theory, high-frequency data is the relative data of daytime or more low-frequency data. High-frequency time series is referred to financial data which is sampled with interval of one hour, one minute, and composed of the price and quantity. Ultra-high-frequency time series is tick-by-tick data in the stock market or in the foreign exchange market。Obviously the ultra-high-frequency data is irregular interval data. After the introduction of theories of (ultra)high-frequency data, we study the fields as follow:1. Engle in 1998 initiatively proposes a new model for the arrival times of trades, termed the autoregressive conditional duration (ACD) model. After Engle's work, several econometricians develop the model. Until now there are more than ten kinds of models. It is difficult of definiting model form and appraising the models because of the inconsistency of model form. This dissertation proposes a Regime-Switching Fractionally Integrated Augmented Autoregressive Conditional Duration (RSFIA-ACD) model. The dissertation also gives the properties of unconditional moment of the model. Estimating the coefficient is developed by the hybrid hierarchy genetic algorithm. The experiment using the Shanghai Stock Market datum illuminates the efficiency of RSFIA-ACD model and the hybrid hierarchy genetic algorithm. RSFIA-ACD model not only includes the appeared models, but also derives dozens of new models.2. High-frequency data is the base of financial market microstructure theory, and interday data can explain the answer which the low-frequency data cannot. This dissertation proves the higher forecasing ability of high-frequency data model by using the theory of VaR and estimating the statistical variable LR.3. Financial time series has the heavy tail, so econometricians want to find the best distribution to analysis the models. This dissertation studys four distributions (Normal, t, general error distribution, SKST), and finds the GED is best one for Shanghai and Shenzhen Stock Markets.