Proximity-based equivalence classes in fuzzy relational database model
share
One of the first attempts to set a solid theoretical foundation for extending the content of relational databases with incomplete information was the fuzzy relational model by Buckles and Petry. This structure was based on two generalizations of the traditional relational model: (1) A tuple component can be any subset of the corresponding domain, rather than a single element and (2) A similarity relation is defined on each domain. This relation satisfies the properties of reflexivity, symmetry and max-min transitivity, thus having the equality relation as a special case. This generalization keeps two key properties of the relational model - that no two different tuples represent the same information and that the application of any operation of the relation algebra has a unique result. Shenoi and Melton generalized this model and showed how the existence of equivalence classes over the attribute domains can also be preserved with a relation that only satisfies the properties of reflexivity and symmetry (proximity relation). The motivation for this generalization is the strictness of the max-min transitivity property of similarity relations, which complicates the construction of this relation for some domain types. An important characteristic of the Shenoi-Melton model is the dependence of the equivalence classes upon the current content of the database. This characteristic, together with the way the equivalence relation is constructed by the proximity relation, can lead to the equivalence classes that don't correspond well with some database query types. Here we will present a different way of forming the equivalence classes over the attribute domains in fuzzy relational databases in which they depend only on the attribute domain and not on the current database state. We will also show a simple method for automatic construction of proximity relations over some domain types.
READ FULL TEXT