Using Loose and Tight Bounds to Mine Frequent Itemsets

Mining frequent itemsets forms a core operation in many data mining problems. The operation, however, is data intensive and produces a large output. Furthermore, we also have to scan the database many times. In this paper, we propose to use loose and tight bounds to mine frequent itemsets. We use loose bounds to remove the candidate itemsets whose support cannot satisfy the preset threshold. Then, we find whether we can determine the frequency of the remainder candidate itemsets with the tight bounds. According to the itemsets that cannot be treated, we scan the database for them. Using this new method, we can decrease not only the candidate frequent itemsets have to be tested, but also the database scan times.