Discovering Injective Mapping Between Relations in Astrophysics Databases

International audienceData in Astrophysics are very often structured with the rela-tional data model. One particularity is that every value is a real number and comes with an associated error measure, leading to a numerical interval [value − error, value + error]. Such Astrophysics databases can be seen as interval-based numerical databases. Classical data mining approach, specifically those related to integrity constraints, are likely to produce useless results on such databases, as the strict equality is very unlikely to give meaningful results. In this paper, we revisit a well-known problem, based on unary inclusion dependency discovery, to match the particularities of Astrophysics Databases. We propose to discover injective mapping between attributes of a source relation and a target relation. At first, we define two notions of inclusion between intervals. Then, we adapt a condensed representation proposed in [15] allowing to find a mapping function between the source and the target. The proposition has been implemented and several experiments have been conducted on both real-life and synthetic databases