2000ǯ 74()     151640ʬ
ֱ   ()

	Nonlinear time series modeling with respect to the trade off 
	between prediction and dynamics

    As one of the most popular topics of common interest during recent
decades, nonlinear time series modeling has attracted great attention from
different disciplines, types of nonlinear model from different theoretical
background have been so far developed by statisticians, dynamists, and
computer scientists. This makes us wonder which one should be used to cope
with our specific nonlinear problems. In the first of this talk, with
respect to the subject of model evaluation, we will give discussion by
reviewing some works on model construction and their applications, from
perspective of prediction and dynamics extraction, to show the reasonability
and limitations of some existing nonlinear models. Then, we will focus on
modeling short-term interest rate being of current interest in the field of
finance engineering. We will introduce a flexible parametric model that we
have recently developed by considering the potential dynamics in interest
rate processes. Empirical results will be given by the comparison between
our model and other interest rate models by using LIBOR data and other
benchmark interest rates, all are involved in developing a model of balance
between predictability and interpretability (dynamics).