高效用项集挖掘算法  

Improved algorithm for mining high utility itemsets

作　　者：祝孔涛[1] 李兴建[2] 王乐[3] 

机构地区：[1]南阳理工学院软件学院,河南南阳473000 [2]南阳理工学院图书馆,河南南阳473000 [3]大连理工大学创新实验学院,辽宁大连116624

出　　处：《计算机工程与设计》2013年第12期4220-4225,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61173163)

摘　　要：现有高效用项集挖掘算法主要采用项集枚举和两阶段方法 (或称为候选项集测试方法),后者时空效率的主要瓶颈在于候选项集过多。针对该问题,采用降低候选项集twu值,及利用项最大最小效用值来估计候选项集的效用值等策略,进而进行剪枝,从而有效地提高挖掘算法的时间和空间效率。实验采用真实数据集和合成数据集进行算法性能测试,实验结果表明改进后的算法性能得到较大的提高。Current approaches on high utility itemset mining are based on either itemset iteration or two-phase （also called candi date itemset testing）, the performance bottleneck of the latter lies on too many candidates it generates. To address this issue, improved strategies is proposed to prune candidate itemsets, including reducing twu value, and estimating utility values of candidates using maximum ＆ minimum item utility value, so as to efficiently improve the time and space efficiency of the mining algorithm. Both real and synthetic datasets are used in the performance test, and experimental results show that the revised algorithm is valid with significant improvement on performance.

关 键 词：效用 高效用项集 频繁项集 候选项集 数据挖掘 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]