以Box—Cox变换和变结构方法为基础,建立了包括变结构、长记忆、非线性等特征的变结构分整增广ACD模型.利用遗传算法等工具,解决了模型参数估计问题,从而解决了ACD模型的形式设定问题.给出变结构分整增广ACD模型的无条件矩性质,并且推导出数十种ACD模型的扩展模型.
The paper proposes a comprehensive model based on the Box-Cox transformation and regime switching methods. It is termed regime-switching fractionally integrated augmented ACD model, which embraces the properties of regime switching, long memory and non-linear structure. The coefficient of the model is estimated by the hybrid hierarchy genetic algorithm, so the handicap in model specification test is overcome. This paper also gives the moment property of the model and derives dozens of new extended models.