计算机工程与应用 ›› 2015, Vol. 51 ›› Issue (22): 243-251.

• 工程与应用 • 上一篇    下一篇

利用CUDA提高内存数据聚类效能的研究

董丽丽,董  玮,张  翔   

  1. 西安建筑科技大学,西安 710055
  • 出版日期:2015-11-15 发布日期:2015-11-16

Research for memory data clustering efficiency with CUDA

DONG Lili, DONG Wei, ZHANG Xiang   

  1. Xi’an University of Architecture and Technology, Xi’an 710055, China
  • Online:2015-11-15 Published:2015-11-16

摘要: 提出一种新的聚类算法AIK-Means,利用CUDA技术加速算法执行效率,并优化内存方法,可在有限时间内进行多次聚类;将Chameleon层次聚类算法用于解决K-Means算法的初始聚类中心敏感问题;在多次聚类结果中用FP-Tree进行关联分析,提高聚类有效性。将算法应用到某集团心理学MMPI数据测试,实验结果表明AIK-Means算法在执行效率和聚类有效性上具有良好的效果。

关键词: 统一计算设备架构(CUDA), Chameleon层次聚类算法, K-Means聚类算法, 频繁模式树(FP-Tree), 明尼苏达多项人格测验(MMPI)

Abstract: This paper puts forward a new clustering algorithm AIK-Means. Multiple clustering can be executed within the limited time by using the CUDA technology, which is able to accelerate execution efficiency of the algorithm and optimize the memory method. The Chameleon hierarchical cluster algorithm is used to solve the initial clustering centers sensitive issues of the K-Means algorithm. In order to improve the validity of clustering, the FP-Tree is used for correlation analysis in several clustering results. In this paper, the algorithm is applied to the psychology MMPI test data of a group. The experimental results indicate that the AIK-Means algorithm performs well in the execution efficiency and cluster validity.

Key words: Compute Unified Device Architecture(CUDA), Chameleon hierarchical cluster algorithm, K-Means algorithm, Frequent Pattern(FP)-Tree, Minnesota Multiphasic Personality Inventory(MMPI)