An Efficient Incremental Mining Approach Based on IT-Tree

Abstract

The itemset-tidset-tree (IT-tree) is an efficient data structure for association-rule mining. Zaki et al. designed a mining algorithm based on the IT-tree structure, which traversed an IT-tree in a depth-first order, generated itemsets by using the concept of equivalence classes, and computed the support values of itemsets fast by tidset intersection. It, however, needed to process all transactions in a batch way. In this paper, we propose a Pre-FUIT algorithm (Fast-Update algorithm based on the IT-tree structure and the concept of PRE-large itemsets algorithm), which does not only update the IT-tree when new transactions are inserted, but also mine all frequent itemsets easily. Experimental results show the good performance of the proposed algorithm.