Inferring aggregation hierarchies for integration of data marts

Abstract

The problem of integrating heterogeneous data marts is an important problem in building enterprise data warehouses. Specially identifying compatible dimensions is crucial to successful integration. Existing notions of dimension compatibility rely on given and exact dimension hierarchy information being available. In this paper, we propose to infer aggregation hierarchies for dimensions from a database instance and use these inferred aggregation hierarchies for integration of data marts. We formulate the problem of inferring aggregation hierarchies as computing a minimal directed graph from data, and develop algorithms to this end. We extend previous notions of dimension compatibility in terms of inferred aggregation hierarchies. © 2010 Springer-Verlag.