Efficient parallel mining of association rules on shared-memory multiple-processor machine

We consider the problem of parallel mining of association rules on a shared memory multiprocessor system. Two efficient algorithms PSM and HSM are proposed. PSM adopted two powerful candidate set pruning techniques distributed pruning and global pruning to reduce the size of candidates, HSM further utilized an I/O reduction strategy to enhance its performance. We have implemented PSM and HSM on a SGI Power Challenge parallel machine. The performance studies show that PSM and HSM outperform CD-SM, which is a shared memory parallel version of the popular Apriori algorithm.