摘要
信息融合数据源的聚类分析对于信息融合过程的算法管理具有重要意义。本文首先提出一种多粒度模糊聚类方法对采集的数据源样本进行聚类分析,得到样本的聚类树。在此基础上,提出一种基于熵增益率的多粒度连续属性离散化算法,将数据源属性离散化。最终得到聚类结果的特征描述,为今后的规则提取做好准备。实验表明:基于多粒度的数据源分析方法是有效的。
Research on data sources is important for Algorithm Management (AM) in information fusion. A method of multi-granularity fuzzy clustering is used on the samples of data sources, and then a clustering tree is obtained. After that, an algorithm of multi-granularity discretization based on entropy is proposed to make the continuous attributes of data sources discrete. Finally, characters of clustering results are described to be ready for knowledge discovery in AM next. The experiment results show that the method based on multi-granularity is effective.
出处
《华东理工大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2008年第4期598-603,共6页
Journal of East China University of Science and Technology
关键词
信息融合
数据源
多粒度
模糊聚类
离散化
information fusion
data sources
multi-granularity
fuzzy clustering
discretization