FUZZY LOSS LESS DECOMPOSITIONS IN DATABASES

In this paper we deal with the problem of defining projection and join operators in the framework of a fuzzy relational database. We study t he properties of such operators relating to the concept of fuzzy funct ional dependency. There are two main properties to achieve: On the one hand, if a relation r satisfies a dependency between two sets of attr ibutes, X and Y, then we want to preserve this dependency in the fuzzy projection over XY of r. On the other hand, we want to test a fuzzy d ependency just by looking at this projection. Therefore, such a projec tion operator must allow us to store the information conveyed in r int o a separate relation with fewer tuples. In order to define the criter ion for fuzzy dependency, we shall need to extend a resemblance relati on defined for crisp data to the fuzzy case. Finally, we shall introdu ce a fuzzy join operator which will allow us to recover the informatio n given in a relation, from its projections. (C) 1998 Elsevier Science B.V. All rights reserved.