Hybrid Self-Organizing Modeling Systems

Models form the basis of any decision. They are used in di?erent context and for di?erent purposes: for identi?cation, prediction, classi?cation, or control of complex systems. Modeling is done theory-driven by logical-mathematical methods or data-driven based on observational data of the system and some algorithm or software for analyzing this data. Today, this approach is s- marized as Data Mining. There are many Data Mining algorithms known like Arti?cial Neural N- works, Bayesian Networks, Decision Trees, Support Vector Machines. This book focuses on another method: the Group Method of Data Handling. - thoughthismethodologyhasnotyetbeenwellrecognizedintheinternational science community asa verypowerfulmathematicalmodeling andknowledge extraction technology, it has a long history. Developed in 1968bythe Ukrainianscientist A.G. Ivakhnenko it combines the black-box approach and the connectionism of Arti?cial Neural Networks with well-proven Statistical Learning methods and with more behavior- justi?ed elements of inductive self-organization.Over the past 40 years it has been improving and evolving, ?rst by works in the ?eld of what was known in the U.S.A. as Adaptive Learning Networks in the 1970s and 1980s and later by signi?cantcontributions from scientists from Japan, China, Ukraine, Germany. Many papers and books have been published on this modeling technology, the vast majority of them in Ukrainian and Russian language.