STRUCTURAL AND PARAMETRICAL IDENTIFICATION OF LLSQ5MODEL WITHOUT THE SOLUTION OF THE NORMAL EQUATIONS SYSTEM

Abstract

The way of calculation of values of the LLSQmodel, in particular MLR (Multiple Linear Regression), without estimate its parameters is offered. The offered way can be used in algorithms of search of the best somewhat the lin ear description of the process submitted by discrete values, for example, by a time series. The recursion scheme of calculation of parameters of the LLSQmodel, alternative to the solution of the Gauss normal equations system is offered.