1 Introduction
Association rule mining (ARM) [16], [22] is essential in discovering correlations within large datasets. ARM is related to the Frequent Itemset Mining (FIM) problem which determines sets of items (i.e., itemsets) that appear frequently in a dataset. Formally, let \mathbf {I} = \lbrace i_1, i_2, \ldots, i_n\rbrace be a set of n items and let \mathcal {D} = [t_1, t_2, \ldots, t_t] be a dataset of t transactions, where each transaction comprises of a set of items, i.e., t_i \subseteq \mathbf {I}. The support of an itemset X \subseteq \mathbf {I}, denoted by \sigma (X), is the number of transactions t_i that include X (i.e., X \subseteq t_i). Given threshold \sigma _0, FIM finds itemsets X such that \sigma (X) \geq \sigma _0, called \sigma _0-frequent items.