一个新的不需要候选集的挖掘关联规则算法——Relim算法的研究  

Study on a New Mining Association Rule Without Candidate Generation^4/―Relim Algorithm

作　　者：刘喜苹[1] 刘彩苹[2] 谭义红[2] 

机构地区：[1]湖南大学软件学院,湖南长沙410082 [2]湖南大学计算机与通讯学院,湖南长沙410082

出　　处：《计算技术与自动化》2006年第2期81-84,共4页Computing Technology and Automation

摘　　要：Fp-growth算法是当前挖掘频繁项目集算法中速度最快,应用最广,并且不需要候选集的一种挖掘关联规则的算法。但是,Fp-growth算法也存在着算法结构复杂和空间利用率低等缺点。Relim算法是在Fp-growth算法的基础上提出的一种新的不需要候选集的挖掘关联规则算法。它具有算法结构简单,空间利用率高,易于实现等显著优点。本文在详细阐述Relim算法后,对Fp-growth算法和Relim算法的性能进行了分析和比较。结果表明,Relim算法尽管结构简单,但其运行速度与Fp-growth算法相比并不慢,而且当对最小支持度高或者频繁规则比较少的数据集进行挖掘时,Relim算法的运行速度往往比Fp-growth算法要快。Fp- growth algorithm is one of the currently fastest and most popular algorithms for mining association rule without candidate generation. However, it has disadvantages such as complicated data structure and lower space utilization rate. Recursivc elimination （Relim） algorithm, which is proposed by Christian Borgelt in 2004 based on the Fp- growth algorithm, is a new algorithm in mining association rule without candidate generation. The dramatic advantages of Relim algorithm are its simplicity of structure and high space utilization rate. In this paper, the Relim algorithm is described firstly, and then is compared with the Fp - growth algorithm. The comparison results show that, even though the Relim algorithm has simple structure, its speed is not slower than that of the Fp - growth algorithm, even faster when the minimal support is set high or there are fewer frequent item sets.

关 键 词：数据挖掘 关联规则 候选集 Fp—growth Relim 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]