Rainer Knauf, Technical University of Ilmenau; Avelino J. Gonzalez, University of Central Florida; Klaus P. Jantke, Hokkaido University
The paper is an attempt to apply the authors’ experiences of their work in validation of rule-based systems to case-based systems. The objective of this work is both to come up with a framework for validation of case-based systems and to answer the question how the knowledge representation and the problem solving paradigm influences the validation technology of an AI system. It turns out, that the general steps of test case validation (test case generation, test case experimentation, evaluation, validity assessment, and system refinement), which are performed in cycles, can and should be the same for both paradigms, but the realization of the particular steps seems to be different. Most of the steps can be performed with less expenditure and less human support for case-based systems than for rule-based systems. Generally, the more a knowledge representation level is explicit and abstract, the more the validation of the system seems to be complicated.