Geracao de Modelos de Regras de Decisap: Uma Abordagem Centrada na Aprendizagem Indutiva (Generation Models of Decision Rules: A Central Approach to Inductive Learning)

摘要

Research for the creation of a mechanism to allow the generation of models (descriptions) of the rules which constitute a knowledge base is presented. The mechanism which was defined is based on an inductive learning process which is proposed, and works from rule classes defined in the base. Such classes are composed of rules which have some common conclusion clause and refer to a same set of objects. The learning approach allows the creation of disjuctive and conjunctive concepts. This process allows the generalization of disjunctive concepts for another that encompasses the former, which represents a simple way of knowledge-based learning. The learning mechanism was tested only for generation of characteristic descriptions, but, as it tried to show, it could also be used for the generation of discriminant descriptions. The results obtained, as well as the acquired experience, allow the conclusion that, in order that the models may better specify the knowledge expressed in the base, more sophisticated ways of defining the rule classes are necessary.