基于ANN－GARCH模型的中国股市收益非对称波动拟合能力研究  

Empirical Study of China’s Security Market Return Volatility:Based on ANN-GARCH Model

作　　者：丁岚[1] 苏治[2] 

机构地区：[1]对外经济贸易大学统计学院北京100029 [2]中央财经大学统计学院,北京100081

出　　处：《统计与信息论坛》2013年第7期22-26,共5页Journal of Statistics and Information

基　　金：对外经济贸易大学校级科研课题《基于贝叶斯理论的金融市场VAR测度-模型与实证》(11YBTJX01)

摘　　要：在传统的非对称GARCH模型上加入ANN逻辑项,从而提高了模型的非对称性描述能力。不同于"黑箱"式的神经网络计算,这种方法的ANN逻辑项是可见、可分析的。通过对上证综指、深证综指和恒生指数的实证研究,发现三个市场都存在"杠杆效应"。研究表明:ANN-GARCH模型总体上比传统模型的拟合效果更好,预测能力更强。This paper will introduce the ANN--GARCH model which adds an ANN logistic component onto traditional asymmetric GARCH models, hoping to improve the capacity of describing the asymmetry. Being different from an ANN computing as ＂black box＂, the ANN logistic addition is visible thus could be analyzed by this new method. Through an empirical study on the Shanghai Securities Composite Index, the Shenzhen Securities Composite Index and the Hang Seng Index of Hong Kong, it presents that these three stock markets all have the ＂leverage effect＂ and that the ANN--GARCH model performs better than the traditional ones, from the fitted capacity to the forecast ability.

关 键 词：收益率波动性 非对称GARCH模型 ANN-GARCH模型 

分 类 号：F830[经济管理—金融学]