一种新的动态关联规则及其挖掘算法  

A new kind of dynamic association rule and its mining algorithms

作　　者：沈斌[1,2] 姚敏[2] 

机构地区：[1]浙江大学宁波理工学院,浙江宁波315100 [2]浙江大学计算机学院,杭州310027

出　　处：《控制与决策》2009年第9期1310-1315,共6页Control and Decision

基　　金：国家自然科学基金项目(60533040;60525202);浙江省自然科学基金重点项目(Z104267);浙江大学宁波理工学院科研启动基金项目

摘　　要：在分析原有定义不足的基础上,提出一种新的动态关联规则,其支持度向量和置信度向量与经典定义相吻合,能更好地反映规则随时间变化的动态信息.进一步提出两种新的动态关联规则挖掘算法:ITS和EFP-growth.其中:两阶段ITS算法具有较好的可理解性;基于扩展FP-树的EFP-growth算法适宜于高密度海量数据的挖掘.实验结果表明,该算法具有较好的挖掘性能和可扩展性,适用于动态关联规则的有效挖掘.In order to improve the shortcomings of the original dynamic association rule （DAR for short）, this paper presents a new kind of DAR, which can reflect the dynamic information of rule with time better. Its defintions of support vector and confidence vector are accord with the classical defintions of support and confidence. Two new mining algorithms are also provided, which are ITS algorithm and EFP-growth algorithm. ITS algorithm has two stages, and can be well comprehended. EFP-growth algorithm is based on extended FP-tree and suitable for large volumes of data with high density. The experiment results show that the proposed algorithm has good performance and is fit for mining DARs.

关 键 词：动态关联规则 扩展FP-树 频数向量 挖掘算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]