Frequent pattern-growth algorithm on multi-core CPU and GPU processors

Abstract

Discovering association rules that identify relationships among sets of items is an important problem in data mining. It's a two steps process, the first step finds all frequent itemsets and the second one constructs association rules from these frequent sets. Finding frequent itemsets is computationally the most expensive step in association rules discovery algorithms. Utilizing parallel architectures has been a viable means for improving FIM algorithms performance. We present two FP-growth implementations that take advantage of multi-core processors and utilize new generation Graphic Processing Units (GPU).