基于项目属性的相联规则提取  

An Algorithm for Discovering Association Rules Based on the Attributes of Items

作　　者：李雄飞[1] 苑森淼[1] 王爱军[1] 郇丹丹[1] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130025

出　　处：《计算机学报》2002年第12期1421-1427,共7页Chinese Journal of Computers

基　　金：国家自然科学基金 ( 6 98730 19);吉林省自然科学基金 ( 19990 5 2 8)资助

摘　　要：相联规则是数据库知识发现领域的重要方法之一 ,用于发现满足用户指定最小支持度和最小信任度阈值的规则 .其中 ,最小支持度阈值确定了研究数据集的规模 ,最小信任度阈值用来衡量一个规则可靠性 .在通常的支持度 /信任度框架下 ,用户只能给出一对最小支持度和最小信任度阈值 ,因此 ,对于所有数据项均采用统一标准处理 .但是 ,实际数据库中的数据项目具有各自的特点 ,该文旨在根据项目的属性特征 ,通过模糊综合评判 ,决定项目合理的最小支持度阈值 ,进而确定各个项目的支持度区间 ,达到在一次数据挖掘中同时发现频繁规则和稀有规则的目的 .由于基于最小信任度的规则提取具有冗余性 ,文中提出规则前件和后件的重要程度对比的思想 ,借助主观判断去除冗余规则 ,从而挖掘出尽可能接近自然的完全规则 .Association rule mining is an important model in data mining. It can find the useful rules according to the thresholds of minsupport and minconfidence that are specified by the user. The minsupport decides the size of the dataset for study, and the minconfidence is used to evaluate the confidence of the rules. But the custom Association Rules Discovery based on the frame of support and confidence has only one threshold of minsupport and minconfidence. It means all the attributes of items in the database are not distinguished. In fact, every item has its own characters. If some rules want to include the rare items, the custom methods will have a dilemma. In this paper, we present a new algorithm, which uses the fuzzy evaluation to decide the reasonable minsupport for each item. Then, the support scopes are generated by cluster. So the different support thresholds can be used in correspond support scopes. We present a notion of importance of rule, which is based on the minconfidence. It can describe that a rule is useful or not. The redundant rules are pruned according to their importance. The results of experiments show that our algorithm is better and more useful.

关 键 词：项目属性 相联规则提取 重要度 频度 对比度 支持度区间 超市 数据库 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]