Knowledge Acquisition Combining Analytical and Empirrcal Techniques

Mario Martín, Ramon Sangüesa, Ulises Cortés

Abstract: In this paper we introduce a methodology for classification-oriented knowledge-base generation using LINNEO, a software for fuzzy classification and rule generation which resorts to analytical -EBG- and empirical -SBL- knowledge acquisition techniques. LINNEO builds a classification from a set of (frequently noisy) observations and a (possibly incomplete) domain theory supplied by the expert. The final result is a fuzzy rules knowledge base. It is believed that integrating both types (EBG, SBL) of learning techniques improves the whole process of knowledge acquistion. In our approach SBL is biased by a domain theory which helps in focusing the induction process.