High-frequency Financial Time Series Modeling,

In recent years, the analysis of financial high frequency time series has been one of the key and difficult issues in the financial community and research circle.We present two different frameworks to analyzing the high frequency time series: first, regularly spaced sampled observations, which is sampled with interval of one hour, one minute even one second; Secondly, the irregularly spaced data, such as transaction by transaction data。The main work and innovations of the dissertation include:1.Through empirical research of high-frequency time series of Shanghai Composite Index, the paper researches the different statistical properties on different frequency.2.The paper studies calendar effects of the sample, such as periodicity and long memory。So we use the flexible Fourier form regression proves and filters the periodic components。And use GHP method to estimate the fractional integration d.3.Introduce various ACD models。Forecasting ability of several parameterizations of ACD models are compared。We use ACD model to study transaction-by-transaction data of China stock market.4.We research SCD model, which is an important extend of ACD model, compare the statistical properties with ACD model deeply.5.The paper builds a new econometric model for estimating both the returns and durations, as well as gives the joint density of the marked point process of duration and transaction-by-transaction returns with an ACD-ARMA framework.