摘要
关联规则挖掘是数据挖掘和知识发现中的一个重要问题,自提出以来得到了广泛的研究。目前关联规则挖掘算法可以分为广度优先算法和深度优先算法两大类,每类都有经典高效的算法提出。但是,这些算法大都是从其自身的角度来描述的,缺乏系统的分类和比较。文章从关联规则挖掘的形式化定义出发,给出频集挖掘的解空间,对两大类算法中的几种经典算法进行了概述,并分析了它们的优缺点。
Association rule mining is an important problem in data mining and KDD, which has been researched widely. By far, association rule mining algorithms can be divided into two main classes: width first and depth first. There are classical and efficient algorithms in each class. But, these algorithms are more or less described on their own, and systematic classification and comparison are absent. This paper begins with the formal definition of association rule mining, introduces the result lattice of frequent itemset mining, summarizes several classical algorithms of the two class, and analyses the advantages and disadvantages of them.
出处
《微电子学与计算机》
CSCD
北大核心
2005年第6期68-72,共5页
Microelectronics & Computer
基金
国家863计划项目(2002AA104240)
中国科学院"十五"重大项目(INF105-SDB)
关键词
数据挖掘
关联规则
频集
等价类
Data mining, Association rule, Frequent itemset, Equivalence class