Accuracy and compactness criteria for evaluating the quality of fuzzy knowledge bases in identification problems

Abstract

The knowledge base quality is viewed as the property of the model of dependence under study to satisfy customer`s requirements according to many criteria, accuracy and compactness being the most popular among them. Fuzzy knowledge bases that simulate three types of dependences are investigated with the output in the form of a crisp number, fuzzy number or a class of solutions. For each type of the dependences accuracy criteria of corresponding knowledge bases are systemized. For evaluating the compactness of a fuzzy knowledge base 9 known partial criteria are described and 5 new criteria are proposed.