8th Information Processing and Management of Uncertainty in Knowledge-Based Systems - IPMU 2000, Madrid (Spain). 03-07 July 2000
Summary:
In this paper we propose a new inductive method to estimate fuzzy rules directly from data when there is no a priori knowledge about the fuzzy rules and/or input fuzzy sets. This method is based on the ORTHO model, an efficient approach to flexible and robust surface fitting under stringent high noise conditions which main strengths are interpretability and computational efficiency. The approach consists in building an ORTHO model and then creating the set of fuzzy rules by interpreting the model. This is stated after establishing the relationship between this automatic learning model and existing neurofuzzy models. Two types of rules are taken into account: Takagi-Sugeno and Zadeh-Mamdani fuzzy rules. The proposed technique is assessed on a synthetic test problem, closer to real world data.
Keywords: fuzzy-rules induction, automatic learning, regression modeling, data mining, rule extraction
Publication date: 2000-07-03.
Citation:
E.F. Sánchez-Úbeda, L. Wehenkel, Automatic fuzzy-rules induction by using the ORTHO model, 8th Information Processing and Management of Uncertainty in Knowledge-Based Systems - IPMU 2000, Madrid (Spain). 03-07 July 2000.
