An analysis of the models informativeness in parametric identification problems

Abstract

Abstract Solving problems of parametric identification with uncertainty in the initial data due to their inaccuracy and incompleteness complicates, and, in some cases, even rules out the possibility of using classic mathematical and statistical methods as the latter are meant to deal with sufficiently large amounts of observations. These conditions involve the use of specific methods to directly determine the exact type of functional link and to verify its qualitative characteristics. A set of such characteristics also forms an idea of the informativeness of a model. This paper considers the characterization of this concept in relation to parametric identification problems and presents a method of their solution where the procedure of the analysis of informativeness of the obtained models is formalized. Approbation of the developed technique was carried out on the example of parametric identification of the time series model. This time series cannot be linearized. A small number of observations were also known. All these conditions precluded the use of statistical methods to construct a time series model.