基于模糊列表的相关高模糊效用挖掘算法
Correlated High Fuzzy Utility Mining Algorithm Based on Fuzzy List
王斌 1李晓华 1周伟 1胡克勇1
作者信息
- 1. 青岛理工大学信息与控制工程学院,山东 青岛 266000
- 折叠
摘要
针对传统的两阶段高模糊效用挖掘算法存在产生大量候选项集、忽略项集之间的联系和需要重复扫描数据库的问题,提出了一阶段基于模糊列表的相关高模糊效用挖掘算法(Correlated High Fuzzy Utility Mining Algorithm Based on Fuzzy List,CoHFUIM).算法设计了新的模糊列表结构(FHUI-list),使挖掘过程仅需扫描一次数据库,提高了运行效率;上述算法增加了相关性约束并提出了Cos-prune剪枝策略,减少了候选项集的数量,使挖掘出的项集既是高效用的也是高相关的;为了使上述算法适用于动态数据库,提出了改进算法CoHFUIM+.在Chess、Connect和Mushroom三个真实数据集进行仿真,结果表明改进算法的运行时间、内存使用及延展性均优于经典算法TPFU.
Abstract
This paper proposes a one-phase correlated high fuzzy utility mining algorithm based on fuzzy list(Co-HFUIM)to address the problem that the traditional two-phase high fuzzy utility mining algorithm produces a large number of candidate itemsets,ignores the relationship between itemsets,and requires scanning the database repeatedly.The algorithm designs a new structure of fuzzy list(FHUI-list)to scan the database only once in the min-ing process,which improves the operation efficiency.The algorithm adds the correlation constraint and the proposed Cos-prune strategy,which reduces the number of candidate itemsets and makes the mined algorithm both efficient and highly correlated.An improved algorithm CoHFUIM + is proposed to apply to dynamic database.Simulation experiments on three real datasets of Chess,Connect and Mushroom show that the algorithm is superior to TPFU in running time,memory usage and ductility.
关键词
相关性/模糊列表/动态收益/效用挖掘Key words
Correlation/Fuzzy list/Dynamic profit/Utility mining引用本文复制引用
出版年
2024
NETL
NSTL
维普
万方数据