On the comparison of imprecise values in fuzzy databases

A comparison measure is proposed that allows to evaluate the extent to which two possibility distributions are interchangeable. This measure is based on two resemblance relations: one expressing resemblance between crisp values of a given domain, the other expressing approximate equality between possibility degrees. The proposed measure allows to compute a degree of synonymy between two representations of imprecise values and generalizes the underlying resemblance relation used to compare crisp values. This interchangeability measure can be used to perform approximate data reduction on fuzzy relations as well as to express new types of fuzzy queries