Improvement on the Constrained Association Rule Mining Algorithm of Separate

The problem of constrained association rule mining in large databases has been attached special research attention, and several algorithms have been introduced in recent years. Separate is a desirable algorithm in terms of efficiency and candidate generation. However, Separate is not perfect due to deficiency of its joint function, especially when the length of itemset or the number of candidate itemsets is large. In this paper, three lemmas are proposed and proved mathematically; and based on these lemmas, a novel early stop function is designed elaborately. The early stop algorithm is capable of breaking the process of loop in the case of dissatisfying the join term, and by this means, performance is improved remarkably. Experiments have demonstrated that the proposed algorithm is more preferable compared with the currently-used join function.