Les Dépendances Difonctionnelles Floues

Abstract

Jaoua and al. proposed an application of difunctional relations in the relational data model, namely difunctional dependencies. They proposed a set of inference rules for functional dependencies and a binary decomposition approach of a relation without loss of information. Integrity constraints, especially fuzzy functional and multivalued dependencies, have been widely studied in various extended relational data models. In this paper, we propose an extension of difunctional dependencies in the framework of fuzzy relational database, in which every fuzzy relation is a set of weighted tuples. We define and characterize the fuzzy difunctionality concept, and we show that inference rules applied in the classic case remain valid in the case of fuzzy relations. Then, we introduce a hierarchical decomposition approach of a fuzzy relation which allows a better clustering of data and reduces their storage space by a particular inheritance mechanism.