Normative Forecasting and Optimal Control of Multidimensional Objects Using a Self-Organization of a System of Non-Physical Models
Alexey G. Ivakhnenko, Grigoriy A. Ivakhnenko · Journal of Automation and Information Sciences · 1997
Suppose the input data contain variables averaged over some time interval corrupted with noise. Conventional balance relations (physical models) are overfitted in such condition. In calculating optimal values of variables, non-physical models obtained using the group method of data handling (GMDH) are more accurate. The first example demonstrates the connection of the method with linear programming; the second one demonstrates the application of non-physical models.