东北大学学报(自然科学版) ›› 2007, Vol. 28 ›› Issue (8): 1101-1104.DOI: -

• 论著 • 上一篇    下一篇

一种基于网格的多策略关联规则挖掘算法

王刚;温涛;郭权;马学彬;   

  1. 东北大学软件中心;东北大学软件中心;东软信息学院计算机科学与技术系;东北大学软件中心 辽宁大连116023;辽宁大连116023;辽宁大连116023;辽宁大连116023
  • 收稿日期:2013-06-24 修回日期:2013-06-24 出版日期:2007-08-15 发布日期:2013-06-24
  • 通讯作者: Wang, G.
  • 作者简介:-
  • 基金资助:
    国家高技术研究发展计划项目(2004AA113020)

Grid-based multi-strategy algorithm for association rules mining

Wang, Gang (1); Wen, Tao (1); Guo, Quan (2); Ma, Xue-Bin (1)   

  1. (1) Software Center, Northeastern University, Dalian 116023, China; (2) Department of Computer Science and Technology, Neusoft Institute of Information, Dalian 116023, China
  • Received:2013-06-24 Revised:2013-06-24 Online:2007-08-15 Published:2013-06-24
  • Contact: Wang, G.
  • About author:-
  • Supported by:
    -

摘要: 在综合研究知识网格的基础上,针对网格平台异构性、动态性和分布性等特点提出一种适应网格平台的多策略关联规则挖掘算法(multiple-strategies based hybrid distribution algorithm,MBHD).MBHD在HD算法的基础上,融合了市场机制优化策略、时间戳机制和令牌环策略,有效克服了网格中制约算法性能的负载均衡问题,并避免了不必要的网络负载和网络竞争的产生,从而解决了关联规则挖掘中最关键的算法耗时问题,并提高了算法的效率和性能.实验给出了对应于不同处理器节点数量和最小支持度时的算法响应时间,表明本算法能够很好地适应网格平台特性,且其性能和可扩展性有明显提...

关键词: 关联规则挖掘, 知识网格, 市场经济模型, 时间戳机制, 令牌环策略, 优化算法

Abstract: Knowledge grid is analyzed comprehensively and then a grid-based MBHD (multi-strategy based hybrid distribution) algorithm is put forward in view of the heterogeneous, dynamic and distributive characteristics of grid. The optimization policy of market mechanism is integrated with timestamp mechanism and token-ring policy in MBHD algorithm to solve efficiently the problem of network load balance due to the restricted algorithms in grid and avoid the formation of unnecessary network load and competition. The time spending as a problem crucial to the algorithm for association rules mining is thus solved with the efficiency and performance improved. Testing results show the MBHD response time performance corresponding to different numbers of processor nodes and minimum support, which is compared with CD, IDD and HD algorithms. A conclusion is reached that the MBHD algorithm is highly adaptable to the characteristics of grid as a platform with its performance especially the expandability improved greatly.

中图分类号: