摘要
为了提高K-prototypes算法的聚类准确度,解决其随机选取聚类中心初始值导至的聚类精度较低和聚类结果不稳定的问题。通过对混合属性数据聚类算法的研究,对K-prototypes算法做了进一步的改进。提出了混合属性聚类的初始聚类中心确定方法,并且通过加权算法改进了相异度计算公式。最后用UCI数据集对算法进行检验,结果表明,改进的加权K-prototype算法更加稳定,并具有较高的聚类精度。
In order to enhance the precision of K-prototypes algorithm and solve the problem that selecting clustering center randomly can reduce the clustering accuracy and lessen clustering stability.Through the research of clustering methods for mixed at- tributes data, some improvements of K-prototypes have been put forward, we proposed a method which can identify the clustering center of mixed attributes data.We also modified the dissimilarity degree formula with weighted algorithm and tested the algorithm with UCI data. The result shows that the enhanced weighted K-prototypes algorithm is better in precision and stability.
出处
《激光杂志》
CAS
CSCD
北大核心
2014年第1期18-20,共3页
Laser Journal
基金
中国移动新疆分公司研究发展基金项目(项目编号:xjm2011-1)
关键词
数据挖掘
混合属性
聚类
权重调整
data mining
mixed attributes
clustering
weight adapt K-prototype