摘要： 提出了一个基于定向原子规则的非精确关联规则挖掘算法 ARA,以支持用户兴趣导向的探索式知识发现。ARA 算法采用一个两层的 Hash表结构对数据库事务中与定向原子规则对应的项进行计数。复合规则通过频繁的原子规则前件项组合和支持度和置信度的估算得到。ARA 只需一次数据库扫描。实验结果表明 ARA 算法的速度快,消耗的内存少,非常适合大型数据库和响应速度要求高的数据挖掘环境。

关键词: 知识发现, 关联规则, 数据挖掘, 原子规则, 复合规则

Abstract: Auser interest-directed, non-accurate and atomic rule-based algorithm (ARA) for association rule mining is proposed for exploratory knowledge discovery. ARA uses a 2-tier Hash table structure for counting the atomic-rule-related items occurring in database transaction. The compound rules are generated by combination among the frequent antecedent items of atomic rules and the estimation of support and confidence for compound rules. ARA requires only one pass of database. The proposed ARA algorithm is very fast and needs low memory consumption. So it’s very suitable for data mining on large database and under the environment in demands of rapid response.

中图分类号: 

• TP311.13