改进的基于距离的关联规则聚类  

Improved distance-based clustering of association rules

作　　者：田宏[1] 王亚伟[1] 王毅[1] 

机构地区：[1]大连交通大学软件学院,辽宁大连116028

出　　处：《计算机工程与设计》2009年第5期1204-1206,共3页Computer Engineering and Design

基　　金：辽宁省教育厅计划基金项目(2008093)。

摘　　要：关联规则挖掘会产生大量的规则,为了从这些规则中识别出有用的信息,需要对规则进行有效的分类组织。现有的规则聚类方法往往直接计算规则间的距离,忽略了项与项之间的联系,不能精确得出规则间的距离。提出一种改进的规则间距离的度量方法,首先计算项间的距离,其次计算相集间的距离和规则间的距离,最后基于此距离利用DBSCAN算法对关联规则进行聚类。实验结果表明,此方法是有效可行的,并能准确发现孤立规则。Large quantities of rules are produced by association rule mining. In order to identify valuable information from these association rules, these rules have to be structured effectively. Since most of existing methods compute distance between rules directly, the correlations hidden in these items are neglected, and then exact distance between rules cannot be obtained. An improved distance metric approach between rules is proposed. First, distance between items is computed. Second, distance between itemsets and between rules is computed. Last, these rules by DBSCAN algorithm are effective and can discover outliers accurately. clustered. Experimental result shows that the new approach is feasible,

关 键 词：关联规则 聚类 项集 距离 基于密度的聚类算法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]