关键词: 数据挖掘, 关联规则, 频繁项集, 增量更新

Abstract: Incremental updating of frequent itemsets in a database includes three problems. An algorithm FIU is proposed for the three problems. Firstly, FP-tree which save transaction recorders of database is materialized on disk. Secondly, When frequent itemsets need be found with new support threshold, only work is to read FP-tree to memory from disk, then travel the tree to find itemsets under new support threshold. Thirdly, when new data insert into database, a new item table of database need be build, then a new FP-tree is built based on the new item table. All transaction records, which were saved in FP-tree on disk, are acquired, then insert into the new FP-tree. Finally, updated frequent itemsets may be found in the FP-tree.

Key words: data mining, association rules, frequent itemset, incremental updating