分布式全局最大频繁项集挖掘算法
来源期刊：中南大学学报(自然科学版)2012年第9期
论文作者：杨君锐 何洪德 杨莉 李海文 薛萍
文章页码：3517 - 3523
关键词：数据挖掘；关联规则；分布式挖掘；最大频繁项集
Key words：data mining; association rules; distributed mining; maximal frequent itemsets
摘 要：提出一种分布式全局最大频繁项集挖掘算法(DMFI)，该算法含局部挖掘与全局挖掘2个阶段。提出一个基于FP-tree的改进频繁模式树(IFP-tree)来存储数据信息。在局部挖掘阶段，先在各站点上分别建立该模式树，并使用有序方式存储频繁项目，然后，通过对各局部数据库的扫描，挖掘出局部最大频繁项集。在全局挖掘阶段，利用各局部数据库生成的最大频繁项集以及利用组通信播报消息的方式，从而挖掘出全局最大频繁项集的集合。对算法的实现以及在多种情况下进行测试。研究结果表明：DMFI算法具有较好的性能。
Abstract: A new algorithm, named distributed maximal frequent itemsets (DMFI) for mining distributed global maximal frequent itemsets from databases was proposed. DMFI has the local mining phase and the global mining phase. A new frequent pattern tree structure, named improved frequent pattern tree (IFP-tree) based on FP-tree, was developed to facilitate the storage. During the local mining phase, DMFI firstly created the tree on each node and used figure sequence to store frequent itemsets, then it discovered the local maximal frequent itemsets after scanning the local databases. During the global mining phase, DMFI was used to share with all nodes in the local maximal frequent itemsets and broadcasted itemsets information for sets communication, so that the global maximal frequent itemsets was mined. DMFI was implemented to evaluate its performance for various cases. The results demonstrate better performance than other algorithms.
杨君锐，何洪德，杨莉，李海文，薛萍
(西安科技大学 计算机科学与技术学院，陕西 西安，710054)
摘 要：提出一种分布式全局最大频繁项集挖掘算法(DMFI)，该算法含局部挖掘与全局挖掘2个阶段。提出一个基于FP-tree的改进频繁模式树(IFP-tree)来存储数据信息。在局部挖掘阶段，先在各站点上分别建立该模式树，并使用有序方式存储频繁项目，然后，通过对各局部数据库的扫描，挖掘出局部最大频繁项集。在全局挖掘阶段，利用各局部数据库生成的最大频繁项集以及利用组通信播报消息的方式，从而挖掘出全局最大频繁项集的集合。对算法的实现以及在多种情况下进行测试。研究结果表明：DMFI算法具有较好的性能。
关键词：数据挖掘；关联规则；分布式挖掘；最大频繁项集
YANG Jun-rui, HE Hong-de, YANG Li, LI Hai-wen, XUE Ping
(College of Computer Science and Technology, Xi’an University of Science and Technology, Xi’an 710054, China)
Abstract:A new algorithm, named distributed maximal frequent itemsets (DMFI) for mining distributed global maximal frequent itemsets from databases was proposed. DMFI has the local mining phase and the global mining phase. A new frequent pattern tree structure, named improved frequent pattern tree (IFP-tree) based on FP-tree, was developed to facilitate the storage. During the local mining phase, DMFI firstly created the tree on each node and used figure sequence to store frequent itemsets, then it discovered the local maximal frequent itemsets after scanning the local databases. During the global mining phase, DMFI was used to share with all nodes in the local maximal frequent itemsets and broadcasted itemsets information for sets communication, so that the global maximal frequent itemsets was mined. DMFI was implemented to evaluate its performance for various cases. The results demonstrate better performance than other algorithms.
Key words:data mining; association rules; distributed mining; maximal frequent itemsets
