A CHAOS-BASED ANALYSIS OF DISCRIMINATIVE POWER OF FUZZY MODELS

A fractal/chaos analysis and discriminative power evaluation of knowle dge bases are presented. They can be used by knowledge engineers to ma ke knowledge acquisition activities more objective. The fractal analys is gives a numerical parameter (fractal dimension). It describes certa in relations between general and specific knowledge items (for example between on-line measurements and general rules of thumbs). The discri minative power analysis is used to quantify one aspect of a knowledge base ''quality.'' Both formal tools are applicable regardless of types of knowledge bases. The article describes their fuzzy interpretations . A simple demonstrative example (a set of 17 fuzzy statements) and a realistic fractal recommendation (fuzzy versus neural control algorith ms) are presented in detail. (C) 1996 John Wiley and Sons, Inc.