An inference engine based on fuzzy logic for uncertain and imprecise expert reasoning

  • Authors:

  • R. O. D'Aquila;C. Crespo;J. L. Mate;J. Pazos

  • Affiliations:

  • Laboratorio de Inteligencia Artificial, Instituto Tecnológico de Buenos Aires, Avda. Madero 399 (1106), Buenos Aires, Argentina and Escuela Superior Técnica, Cabildo 30, Buenos Aires, Ar ...;Laboratorio de Inteligencia Artificial, Instituto Tecnológico de Buenos Aires, Avda. Madero 399 (1106), Buenos Aires, Argentina;Facultad de Informática, Universidad Politécnica de Madrid, Campus de Montegancedo, s/n 28660 Boadilla del Monte, Madrid, Spain;Facultad de Informática, Universidad Politécnica de Madrid, Campus de Montegancedo, s/n 28660 Boadilla del Monte, Madrid, Spain

  • Venue:
  • Fuzzy Sets and Systems - Data bases and approximate reasoning
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper addresses the development and computational implementation of an inference engine based on a full fuzzy logic, excluding only imprecise quantifiers, for handling uncertainty and imprecision in rule-based expert systems. The logical model exploits some connectives of Lukasiewicz's infinite multi-valued logic and is mainly founded on the work of L.A. Zadeh and J.F. Baldwin. As it is oriented to expert systems, the inference engine was developed to be as knowledge domain independent as possible, while having satisfactory computational efficiency. This is achieved firstly by using the same linguistic term set in every universe of discourse. Thus, it is possible to add a dictionary to the knowledge base, which translates the usual linguistic values of the domain to those of the term set. Secondly, the logical operations of negation and conjunction and the modus ponens rule of inference are implemented exclusively in the truth space.The approach provides, firstly, a realistic and unambiguous solution to the combination of evidence problem and, secondly, offers two alternative versions of implementation. The full version uses the algorithms of the operations involved. In a more efficient version, which places a small constraint on the use of linguistic modifiers and is confined to knowledge bases whose inference chains are no longer than three links, the above algorithms are replaced by pre-computed tables.