Summary-Based Pattern Tableaux Generation for Conditional Functional Dependencies in Distributed Data

Conditional Functional Dependencies (CFD) are an extension of Functional Dependencies (FDs) that capture rules about the data consistency. Existing work on discovering CFDs focused on centralized data. Here, we extend this work to horizontally distributed relations. Given an embedded functional dependency, we generate a pattern tableau that represents a CFD. The original feature of our work is generating CFD pattern tableaux from a distributed relation, without merging all the distributed tuples in a centralized relation. We propose a distributed algorithm based on the concept of pattern summary that minimizes data shipping between the sites of distributed relation.