New Tighter Upper Bounds for Mining High Average-Utility Itemsets
Jimmy Ming‐Tai Wu, Jerry Chun‐Wei Lin, Matin Pirouz, Philippe Fournier‐Viger · 2018
In the past, frequent itemset mining (FIM) revealed the high-frequent patterns but ignored the more important concepts such as unit of profit and quality of the items. Recently, high-utility mining (HUIM) has caused wide public concern in the data mining field. A principal problem in HUIM is that the HUIM needs to handle the exponential search space for mining high-utility itemsets while the number of distinct items and the size of the database are both very large. High average-utility itemset mining (HAUIM) is an extension for traditional HUIM concept to provide a different measure with HUIM. It mines the average-utility value of the itemsets regarding to the length of it. Two new tighter upper-bounds, maximum following utility upper-bound (mfuub) and top-k revised transaction maximum utility upper-bound (krtmuub), are proposed in this article to further contract the size of candidate pattern set. Experiments were conducted on two benchmark datasets to show that the proposed method outperforms the previous HAUIM algorithms in terms of runtime