Abstract:
In this paper, a new method of clustering discrete data is presented. The dependency structure is combined with the Gibbs sampling to cluster. The efficiency of sampling can be markedly improved and the problems resulted from EM algorithm can be avoided. Experimental results show that this method can effectively cluster discrete data.
Key words:
Clustering; Discrete data; Dependency structure; Gibbs sampling; MDL criterion
摘要： 建立了一种新的离散数据聚类方法，该方法结合变量之间的依赖结构和Gibbs sampling 进行离散数据聚类，能够显著提高抽样效率，并且避免使用EM 算法进行聚类所带来的问题。试验结果表明，该方法能够有效地进行离散数据的聚类。
关键词:
聚类；离散数据；依赖结构；Gibbs 抽样；MDL 标准
WANG Shuangcheng, YU Shiquan, CHENG Xinzhang. Clustering of Discrete Data Based on Dependency Structure and Gibbs Sampling[J]. Computer Engineering, 2006, 32(9): 28-30.
王双成，俞时权，程新章. 基于依赖结构和 Gibbs Sampling 的离散数据聚类[J]. 计算机工程, 2006, 32(9): 28-30.