GC-tree: A fast online algorithm for mining frequent closed itemsets

Abstract

Frequent closed itemsets is a complete and condensed representaion for all the frequent itemsets, and it's important to generate non-redundant association rules. It has been studied extensively in data mining research, but most of them are done based on traditional transaction database environment and thus have performance issue under data stream environment. In this paper, a novel approach is proposed to mining closed frequent itemsets over data streams. It is an online algorithm which update frequent closed itemsets incrementally, and can output the current closed frequent itemsets in real time based on users specified thresholds. The experimental evaluation shows that our proposed method is both time and space efficient, compared with the state of art online frequent closed itemsets algorithm FCI-Stream [3]. © Springer-Verlag Berlin Heidelberg 2007.