关键词: 数据挖掘, 频繁模式树, 全局最大频繁项集, 算法

Abstract: Most mining maximum frequent itemsets algorithm based on local data base, so a fast algorithm for Mining Global Maximum Frequent Itemsets based on Frequent pattern tree(MGMFIF) was proposed. MGMFIF mined all global frequent items and made itemset, then local Frequent-Pattern tree (FP-tree) of each node was constructed based on this itemset. Finally, this itemset was chose as global maximum frequent itemsets, and all the global maximum frequent itemsets were obtained by top-down strategy. By adopting FP-tree structure, MGMFIF greatly reduced database scanning times and runtime comparing to Apriori-like algorithms. MGMFIF remarkably lessened candidate itemsets and communication traffic by using top-down strategy. Experimental results suggested that MGMFIF was fast and effective.

Key words: data mining, frequent-pattern tree, global maximum frequent itemsets, algorithms