New Algorithm of Maximum Frequent Itemsets for Mining Multiple-Level Association Rules

Discovering maximum frequent itemsets is a key issue in data mining applications. Most of the previous studies adopt an Apriori-like candidate itemsets generation-and-test approach, however, candidate itemsets generation is costly. In this study, we propose a new algorithm named ML_Pincer for discovering maximum frequent itemsets in multiple-level association rules. ML_Pincer algorithm combines the top-down and the bottom-up directions progressive deepening searching ideas, moreover, it uses two-way pruning tactic: the information which gathered in one direction can prune more candidate itemsets during the search in the other direction. It decreases candidate itemsets greatly and avoids making multiple passes over database, consequently, it reduces CPU time and I/O time remarkably. Experiments prove that ML_Pincer algorithm is more efficient than PMAM algorithm, especially when some maximum frequent itemsets are long.