基于分类信息的关联规则间距离函数的改进  

DISTANCE FUNCTIONS OF ASSOCIATION RULE BASED ON TAXONOMY INFORMATION

作　　者：梁敏[1] 阮备军[1] 朱扬勇[1] 

机构地区：[1]复旦大学计算机与信息技术系,上海200433

出　　处：《计算机应用与软件》2006年第3期7-9,14,共4页Computer Applications and Software

基　　金：国家863高技术基金项目(No.2001AA113181)资助

摘　　要：规则聚类将关联规则挖掘产生的大量规则重新组织,帮助用户发现感兴趣的规则。规则距离函数是规则聚类中的重要一环。本文基于分类信息的层次结构特点,对已有的规则距离函数进行改进,提出了参数控制距离和递归权重距离的方法,解决了以往规则距离函数的多次匹配问题。通过距离函数的对比实验,证明提出的方法是合理有效的。Rule clustering re-organizes rules which are generated during the procedure of association rule mining, to help customers find out interested rules, The rule distance function is the key feature of rule clustering. On the basis of the knowledge on the hierarchy of taxonomy information,this paper presents the methods of Parameters Controlling Distance and Recursive Weights Distance, to deal with the problem of multiple occurrences which is not investigated in the previous research. According to the corresponding comparison experiment of the rule distance functions,the validity and rationality of the proposed methodology can be proved.

关 键 词：距离函数 关联规则 聚类 数据挖掘 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] O431.2[自动化与计算机技术—计算机科学与技术]