挖掘相联规则的并行算法研究  

Parallel Algorithm for Discovering Association Rules

作　　者：臧雪柏[1] 李雄飞[1] 全勃[1] 李军[1] 

机构地区：[1]吉林大学计算机科学与技术学院,吉林长春130025

出　　处：《小型微型计算机系统》2003年第12期2260-2263,共4页Journal of Chinese Computer Systems

基　　金：吉林省应用基础研究基金 (1 9990 52 8和 2 0 0 30 52 2 - 1 )资助

摘　　要：在研究多段支持度数据挖掘算法的基础上提出并行挖掘相联规则的算法 .给出了在并行条件下以负载平衡为目的的种子项集的划分的贪心算法策略 .基于多段支持度特征 ,为减少各个处理机之间的制约 ,提出按事务长度进行数据集划分的方案 ,并具体实现了多段支持度的并行算法 .实验结果表明该算法具有很高的效率 .特别是在双CPU情况下 ,并行算法的效率接近顺序算法的两倍 .如果把算法修正和推广到群集环境下 ,会更一步增加相联规则的实用性 .A parallel algorithm for discovering association rules is presented, after an algorithm based on calculating multi-segment support has been studied. The greedy algorithm, which is used to generate the partition of feed set, can guarantee the balance of load in the system. The data set was divided into partitions by the length of transactions, so the restrictions among CPUs are reduced. The results of experiments show that parallel mining association rules improve the efficiency. Especially, the efficiency of the parallel algorithm is almost twice of serial algorithm when system has dual CPUs.If the algorithm is used in computer cluster, it could improve utility of association rules.

关 键 词：频繁项集 多段支持度 并行算法 负载平衡 数据挖掘 相联规则 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] TP18[自动化与计算机技术—计算机科学与技术]