What You Store is What You Get

Abstract

Recent studies emerged the need for representations of frequent itemsets that allow to estimate supports. Several methods have been proposed that achieve this goal by generating only a subset of all frequent itemsets. In this paper, we propose another approach, that given a minimum support threshold, stores only a small portion of the original database from which the supports of frequent itemsets can be estimated. This representation is especially valuable in case one is interested in frequent itemset of size at least some threshold value. However, we also present methods for estimating the support of smaller frequent itemsets.