Mining frequent patterns from incremental databases

Mining frequent itemsets is an important task in data mining area, which is to find the itemsets frequently purchased together from a transaction database. However, new transactions will be continuously added into the transaction database, and thus the frequent itemsets will be changed with time. Users may eager for getting the latest frequent itemsets from the updated database as soon as possible in order to make the best decision. Therefore, it has become an important issue to work out efficient ways of finding the latest frequent itemsets when the transactions keep being added to the database. Although tree-based approaches have been recently adopted in most of the studies in this field, they have shown the two common problems: repeated adjustments of tree nodes and a need for large amounts of memory space. In order to improve the previous approaches, this paper proposes an efficient algorithm which only keeps frequent itemsets in the tree structure. When a set of new transactions is added, the tree structure can be updated by only adjusting few tree nodes for our algorithm.