与时机判定相结合的关联规则增量更新算法

现有的关联规则更新算法大多致力于解决增量更新本身,但很少同时考虑更新时机,不适于对实时应用中频繁更新的数据进行有效处理.针对此问题,提出了一种与时机判定相结合的关联规则增量更新算法,在改进增量更新方法的同时,兼顾对更新时机的判定.在关联规则增量更新阶段,计算含有非空子集个数之和最多的频繁项集,找出在更新数据集中仍然频繁的项集,根据Apriori性质,避免对其子集的处理,从而实现对候选项集的有效剪枝.实验结果表明,该算法通过对更新时机的及时判定和候选项集的有效剪枝,提高了关联规则的更新效率.

Incremental update algorithm of association rules combining with moment checking

Most of the present update algorithms of association rules are devoted to solve the problem of incremental update, but seldom check the update moment simultaneously. It is unsuitable to deal with the data in real time applications which are updated frequently. An improved incremental update algorithm of association rules is proposed. It includes the moment checking phase and the incremental update phase. In the phase of moment checking, the difference between association rules is calculated to decide whether update or not. During the phase of incremental update, frequent itemsets whose number of non-empty subsets is the largest are acquired firstly, and the pruning of candidate itemsets is effective because the computation of the subsets is unnecessary according to the Apriori property. Experimental results show that the proposed pruning strategy improves the update efficiency of association rules.