A New Fuzzy Inferencing Method for Inductive Reasoning

Francisco Mugica Alvarez
Instituto Latinoamericano de la Comunicación Educativa
Calle del Puente, 45
Col. Ejidos de Huipulco, Delegación Tlalpan
14380, México, D.F.
México
Email: FMugica@ILCE.Edu.MX
François E. Cellier
Institute of Computational Science
ETH Zürich
CH-8092 Zürich
Switzerland
Email: FCellier@Inf.ETHZ.CH

Keywords

Fuzzy Inferencing
Defuzzification
Fuzzy Systems
Fuzzy Control
Inductive Reasoning

Abstract

In this paper, a new fuzzy inferencing method is presented, and its performance is compared to that of other fuzzy inferencing methods (sometimes referred to as "defuzzification" methods) that are frequently cited in the literature. The choice of the fuzzy inferencing engine can decide over success or failure of applications of fuzzy technology in areas such as qualitative modeling or qualitative control. It will be demonstrated that the new fuzzy inferencing technique, called the five-nearest-neighbors (5NN) rule, performs exceedingly well in a data-driven environment, i.e., in a situation where a fuzzy system is automatically being synthesized from available measurement data of its surroundings. All fuzzy inferencing algorithms that are compared in this paper have been implemented in SAPS-II, an experimental software for the automated synthesis of qualitative models, fuzzy controllers, and fuzzy inductive reasoners. The advantages of the new fuzzy inferencing method will be shown by means of the identification of a qualitative abstraction of a differential equation model.

Interested in reading the full paper? (11 pages, 591,661 bytes, pdf)

Homepage