On the Self-organizing Induction-Based Intelligent Modeling

Abstract

The article considers the issues of intellectualization of data-driven means for modeling of complex processes and systems. Some relevant terms of the modeling subject area are analyzed for an adequate explanation of the difference between theory-driven and data-driven approaches. The results are presented of the Internet retrieval for journal and book sources containing the term “intelligent modeling” and its variations in their titles and texts. Analysis of these sources made it possible to suggest an advanced conception of the intelligent modeling. It introduces three main levels of intellectualization of such means: offline intelligent modeling for constructing models of objects from available data; online intelligent modeling in an operating system of control or decision making; comprehensive intelligent modeling of work modes of a complex system. The original features of GMDH-based self-organizing inductive modeling are characterized showing that GMDH is one of the most powerful methods of data mining and computational intelligence for tasks being solved under conditions of uncertain and incomplete prior information. The inductive modeling algorithms can be the reasonable basis for creating advanced intelligent modeling tools.