一种简单有效的并行化频繁项集挖掘算法  

An efficient simple parallel frequent itemset mining algorithm

作　　者：金桃[1] 何艳珊[2] 宋伟国[2] 岳敏[2] 

机构地区：[1]吉林广播电视大学,吉林长春130022 [2]兰州大学信息学院,甘肃兰州730000

出　　处：《微计算机信息》2010年第18期147-149,共3页Control & Automation

摘　　要：频繁项集挖掘,作为数据挖掘的一项基本任务,自提出以来就受到计算机科学理论研究的广泛重视。当前很多数据挖掘算法在处理大型数据集时的运行时间无法预计并且不可接受,这也是当前频繁项集挖掘领域遇到的一个主要挑战。本文提出一种简单有效的并行化频繁项集挖掘算法-SP-FP-Growth算法,此算法可以有效的挖掘频繁项集,并通过极少的节点间通讯和负载平衡策略保证了并行环境的高效性。实验结果证明,SP-FP-Growth算法具有良好的时间效率和可伸缩性。As a basic task of data mining, mining frequent itemset has been widely concerned by computer science research since it was proposed. The problem is many existing data mining algorithm cannot efficient run on big dataset, and it is also the main challenge for frequent itemset mining. In this paper, a new efficient simple parallel frequent itemset mining algorithm SP-FP-Growth is introduced to mine frequent itemset efficiently. The algorithm assures the high-efficiency of parallel environment through very little communication among nodes and load balance strategy. The result of experiment shows that SP-FP-Growth has good performance of time efficiency and scalability.

关 键 词：频繁项集 数据挖掘 并行化 FP—Growth 负载平衡 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]