A concurrent approach to the key-preserving attribute-oriented induction method

Abstract

Attribute-Oriented Induction (AOI) reduces the search space of large data to produce a minimal rule set. Classical AOI techniques only consider attributes that can be generalised but eliminates keys to relations. The Key-Preserving AOI (AOI-KP) preserves keys of the input relation and relate them to the rules for subsequent data queries. Previously, the sequential nature of AOI-KP affected performance on a single processor machine. More significantly, time was spent doing I/O to files linked to each generated rule. AOI-KP is O (np) and storage requirement O (n), where n and p represent the number of input and generalised tuples respectively. We present two enhanced AOI-KP algorithms, concAOI-KP (concurrent AOI-KP) and onLineConcAOIKP of orders O (np) and O (n) respectively. The two algorithms have storage requirement O (p) and O (q), q =p*r, 0