Combining Genetic Algorithms and Case-Based Reasoning for Genetic Learning of a Casebase: A Conceptual Framework

In this paper, we present a conceptual framework that combines genetic algorithms and case-based reasoning (CBR) to first learn a genetic hierarchy of cases and then maintain and refine the casebase as the system runs. We propose to use genetic algorithms to generate useful cases since there is not any actual cases to bootstrap our CBR module. We use these evolved cases to develop and test the various stages of the CBR module such as evaluation and retrieval, adaptation, and learning for refining the module. We propose a fitness measure of a case that is based on not only the combination of its attribute values, but also on it being a member of the casebase, which involves its utility in the CBR module. To promote the synergy between CBR and genetic algorithms for genetic learning, we propose and describe several concepts such as meta-genetic code, evolutionary adjustment, refinement, incompatibility, and breakthrough, population migration, granularization, and injection, and deterministic mating.