An Efficient Method For Finding Emerging Large Itemsets

The incremental mining of association rules has been shown to be more efficient than rerunning standard association rule algorithms such as Apriori. As each increment is processed, we see the emergence of some itemsets. An itemset that has emerged is one that was small and is large in the current increment. An emergent large itemset is a small itemset that has the potential to become large, and will do so with high probability. In this paper we modify an existing incremental algorithm, UWEP, so that it can identify emergent large itemsets. We show that, on average, 65% of the emergent large itemsets identified by the algorithm actually do emerge. General Terms Algorithms, Experimentation