Binary partition based algorithms for mining association rules

Mining association rules is an important data mining problem. A fast binary partition-based algorithm (BPA) for mining association rules in large databases is presented in this paper. Basically, the framework of BPA is similar to that of the algorithm Apriori. In the first pass, all the frequent 1-item sets are divided into two disjoint parts. Accordingly, in each subsequent pass k, we partition the set of all the frequent k-item sets into three subsets. Any two different partitions are disjoint. If necessary, this partitioning procedure can be a recursive one. Therefore, we get a binary partition tree in the first pass and a corresponding ternary partition tree in each subsequent pass k. Due to such a partition, BPA can be very easily parallelized, assuming a shared-memory architecture.