STATISTICAL MODELING OF COMPLEX SYSTEMS

Abstract

The author expounds statistical modeling of complex systems with the use of regression analysis for typical conditions of solution of real applied problems, when the model structure is unknown for the researcher. A necessity of using the experiment design, extended conception of orthogonality, and a system of orthogonal contrasts is shown. The results of modeling the digital balance are presented which have confirmed the expediency of the offered approach and the methods used for obtaining the models. The quality criteria of the obtained model are characterized as the best ones. The model obtained is adequate, highly informative, maximum stable, semantic, all the coefficients are orthogonal to each other. The model use allows increasing 13.3 times the accuracy of the measuring device by the criterion of the average absolute error to 0.012%, and 11.2 times by the criterion of root-mean-square approximation error to 4.80%. The results of the use of the above stated conception of regression analysis have confirmed its efficiency. The method of statistical modeling can be successfully applied in the development of high-tech facilities, high technology, intelligent measuring instruments, machine building, instrument engineering, agrobiology, etc.