基于藤Copula方法的持续期自相依结构估计及预测  

Auto-dependence Structure Estimating and Forecasting of Duration Based on Vine Copula

作　　者：叶五一[1] 李潇颖[1] 缪柏其[1] 

机构地区：[1]中国科学技术大学统计与金融系,安徽合肥230026

出　　处：《中国管理科学》2015年第11期29-38,共10页Chinese Journal of Management Science

基　　金：国家自然科学基金青年面上连续资助项目(71371007);国家自然科学基金面上资助项目(71172214);国家自然科学基金青年科学基金资助项目(71001095)

摘　　要：本文基于Copula方法对由高频分笔数据得到的交易量持续期进行了研究。应用多元藤Copula方法对连续几个交易量持续期之间的自相依结构进行估计,在此基础上提出了一种新的条件密度函数估计方法,进而给出了交易量持续期的预测。对中国石化高频分笔数据进行实证分析的结果表明,本文模型对持续期的预测能力要明显优于EACD模型,在密度函数预测检验方面,本文模型也有更好的表现。In this paper,the trading volume duration sequence derived from high-frequency tick-by-tick data is analyzed by Copula method.The auto-dependence structure of several consecutive trading volume durations is estimated by multivariate vine Copula,then,a new estimating method about conditional density function forecasting is also proposed.Moreover,a new forecasting method of the volume duration is put forward.Empirical results of Sinopec show that the predictive ability of our model is much better than that of EACD,which can also be demonstrated from the density forecasting test.

关 键 词：Canonical藤Copula 自相依结构 ACD模型 高频分笔数据 

分 类 号：C931[经济管理—管理学]