Adaptive time series filters to smooth and forecast economic variables

Four monthly Ukrainian trucking industry time series are investigated within period of January 2003 to February 2007. The AR(1) models are used to filter and predict real values of the series. The problems are solved on the basis of a differentiable matrix functional, defined on the coefficient matrix of the filter. The functional presents an information measure of the distribution parameters of the vector state and its tentative estimations and is well known as Kullback information divergence. (© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim)