Abstract: Aiming at disadvantage of presented top-down mining algorithm that has redundant computing when non-frequent itemsets generate their subsets and prune some repeatedly generated subsets，an algorithm of top-down mining based on orientation subset is proposed，which is suitable for mining relative long frequent itemsets.The algorithm uses orientation subset to generate subsets of non-frequent itemsets by top-down strategy，which can efficiently prune redundant subsets and reduce repeatedly computing to improve its efficiency.Comparing with presented top-down mining algorithms，the experiments indicate that it is fast and efficient.

Key words: orientation subset, top-down, long frequent itemsets, mining algorithm

摘要： 针对现有自顶向下挖掘算法的不足，即在非频繁项目产生子集时和修剪重复产生的子集时存在冗余计算，提出一种基于定位子集的自顶向下挖掘算法，其适合于挖掘较长频繁项目集;算法按自顶向下策略用定位子集的方法产生非频繁项的子集，并有效地修剪冗余子集和减少重复计算，提高了算法的效率。实验证明，与现有的自顶向下挖掘算法相比，该算法是快速而有效的。

关键词: 定位子集, 自顶向下, 长频繁项目集, 挖掘算法