Abstract: This paper considers the problem of losing the rules when people use the Apriori algorithm or the ML-T2 algorithm to mine the association rules in the database.It introduces two feasible solutions. The first: without changing the idea of the algorithm ,from top to bottom,people can make full use of the pre-set minSup. It ensures that the useful rules will all be extracted through pre-keep the possible candidate large itemsets for the low-level at the high-level. The second: it adopts the idea of from bottom to top and without generating the candidate large itemset. It construcs the FP-tree at each level, then people will not lose useful rules.

摘要： 针对Apriori算法及ML-T2算法在应用时会造成规则遗失的问题，文章对关联规则进行了研究，并提供了可行的解决方案。其一为不改变原算法从顶到底的思想，充分利用预先设置的minSup值，通过在高层上预存可能会在底层上构成候选大项集的项集来保证有用规则的完全提取；其二则采用从底到顶的思想，运用不产生候选大项集方法，并把它扩展到由底层到顶层的各层建立各自的FP树，这样也不会造成有用规则的丢失。

关键词: 数据挖掘, 关联规则, 规则遗失

CLC Number: 

• TP311.13