Array-Tree: A persistent data structure to compactly store frequent itemsets

Frequent itemset mining discovers correlations among data items in a transactional dataset. A huge amount of itemsets is often extracted, which is usually hard to process and analyze. The efficient management of the extracted frequent itemsets is still an open research issue. This paper presents a new persistent structure, the Array-Tree, that compactly stores frequent itemsets. It is an array-based structure exploiting both prefix-path sharing and subtree sharing to reduce data replication in the tree, thus increasing its compactness. The Array-Tree can be profitably exploited to efficiently query extracted itemsets by enforcing user-defined item or support constraints. Experiments performed on real and synthetic datasets show both the compactness of the Array-Tree data representation and its efficient support to user queries.