Conference Proceedings

ParaDualMiner: An efficient parallel implementation of the DualMiner algorithm

RMH Ting, J Bailey, K Ramamohanarao, H Dai (ed.), R Srikant (ed.), C Zhang (ed.)

ADVANCES IN KNOWLEDGE DISCOVERY AND DATA MINING, PROCEEDINGS | SPRINGER-VERLAG BERLIN | Published : 2004

Abstract

Constraint based mining finds all itemsets that satisfy a set of predicates. Many constraints can be categorised as being either monotone or antimonotone. Dualminer was the first algorithm that could utilise both classes of constraint simultaneously to prune the search space. In this paper, we present two parallel versions of DualMiner. The ParaDualMiner with Simultaneous Pruning efficiently distributes the task of expensive predicate checking among processors with minimum communication overhead. The ParaDualMiner with Random Polling makes further improvements by employing a dynamic subalgebra partitioning scheme and a better communication mechanism. Our experimental results indicate that bo..

View full abstract