股票市场中的超高频分笔数据分析

发布时间：2018-04-01 16:32

本文选题：超高频数据　切入点：拟极大似然估计　出处：《吉林大学》2014年硕士论文

【摘要】：进入大数据时代，数据的获取与存储已不再成问题，如在股票市场中我们可以获得最精细的Tick数据（即超高频数据）.此类数据几乎包含市场中的所有信息，如何处理这些数据以尽可能获得最大的信息量是我们需要解决的问题. 注意到，Tick数据是非等时间间距的，并且其间距是随机的，处理此类数据有两种方法：一是在将原始数据过滤产生新的等间距的数据（如分钟数据），然后使用经典模型；二是直接对随机时间间隔的超高频数据建立新的模型.第一种处理方法降低了数据频率，相应的要付出信息损失的代价.为了充分利用超高频数据的信息，必须对随机时间间隔进行建模.但是，我们注意到Tick数据并不是完全随机的，表现为时间间隔和价格波动的丛集性（聚类性）效应.为了解决这一问题，本文选用自回归条件持续期模型（ACD模型）来刻画时间间隔的自相关结构，而选用ECOGARCH模型来刻画对数价格波动率的变化，然后将两个模型结合在一起构成ACD-ECOGARCH(1,1)模型来刻画Tick数据.在可观测到所有对数价格跳跃的假设下，我们引入拟极大似然估计方法（QMLE）来对模型中的参数进行估计. 最后将该模型应用于A股市场上海交易所的贵州茅台（证券代码：600519）Tick数据.利用模拟退火算法给出参数估计的数值解，，并对杠杆效应和波动率进行说明.
[Abstract]:In the era of big data, the acquisition and storage of data is no longer a problem. For example, in the stock market, we can obtain the most sophisticated Tick data (that is, ultra-high frequency data). This kind of data contains almost all the information in the market. How to process these data to get the maximum amount of information is the problem we need to solve. Note that the unequal-time interval of the pick data is random. There are two ways to deal with this kind of data: one is to filter the original data to produce new equidistant data (such as minute data), and then to use the classical model; The second is to establish a new model of UHF data with random time interval. The first method reduces the frequency of the data, and the cost of information loss is paid accordingly. In order to make full use of the information of UHF data, It is necessary to model random time intervals. However, we note that Tick data are not completely random, as shown by the clustering effect of time interval and price volatility. In this paper, autoregressive conditional duration model (ACD model) is used to describe the autocorrelation structure of time interval, while ECOGARCH model is used to describe the change of logarithmic price volatility. Then the two models are combined to form the ACD-ECOGARCH1) model to describe the Tick data. Under the assumption that all logarithmic price jumps can be observed, we introduce the quasi-maximum likelihood estimation method (QMLE) to estimate the parameters of the model. Finally, the model is applied to the data of Moutai, Guizhou (stock code: 600519Tick) of Shanghai Stock Exchange. The numerical solution of parameter estimation is given by simulated annealing algorithm, and the leverage effect and volatility are explained.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：O212.1;F830.91

【共引文献】