摘要
利用数据对象间的关联限制可以改善聚类算法的效果,但对于关联限制与K中心点算法的结合策略则少有研究。由此研究了关联限制与PAM算法的结合方法,提出了算法CPAM。首先基于限制找到一个合适的初始分隔;在接下来反复地调整中心点的过程中,也考虑到了所给限制。实验结果显示:CPAM可以有效地利用关联限制来提高一些实际数据集的准确率。
Instance-level constraints have shown to be useful to improve the performance of some existing clustering algorithms.There is yet little research on the approach of instance-level constraint-based K-median algorithm.An instance-level constralnt-based PAM algorithm CPAM is presented.h begins by finding an initial partition that is complying with the constraints.Then it does the replacement of a medoid by a nonmedoid iteratively,at all the time constraints will be taken into consideration.The test on three real datasets suggests that CPAM is effective in utilizing instance-level constraints in clustering some real datasets.
出处
《计算机工程与应用》
CSCD
北大核心
2006年第6期190-192,共3页
Computer Engineering and Applications
关键词
聚类分析
PAM算法
半监督学习
clustering analysis,PAM algorithm,semi-supervised learning
