Graded data dependencies in relational databases: Reasoning with similarity of time

Abstract

We introduce graded if-then dependencies in relational similarity-based databases which formalize functional dependencies in data which is recorded in different time points or intervals. The dependencies formalize rules expressing that similar values of given attributes in similar time points or intervals yield similar values of other attributes. We show how the rules may be formalized in a relational similarity-based model of data, we present basic properties of the rules, and outline methods for computing non-redundant sets of dependencies which characterize all dependencies valid in given data. This paper is a short report on research in progress, proofs of theorems are only outlined or omitted.