Database dependency discovery: a machine learning approach

摘要: Database dependencies, such as functional and multivalued express the presence of structure in database relations, that can be utilised design process. The discovery dependencies viewed an induction problem, which general rules (dependencies) are obtained from specific facts (the relation). This viewpoint has advantage abstracting away much possible particulars dependencies. algorithms this paper designed they easily generalised to other kinds dependencies.Like current approaches computational inductive logic programming, we distinguish between top-down bottom-up algorithms. In a approach, hypotheses generated systematic way then tested against given relation. relation is inspected order see what it may satisfy or violate. We give simple (but inefficient) algorithm, bi-directional algorithm. case these have been implemented FDEP system evaluated experimentally. algorithm most efficient three, also outperforms literature.