多属性加权的模糊c均值聚类算法  

Fuzzy C-means Clustering Based on the Multivariable Weighted Attributes

作　　者：李诗瑾 李倩[1] 徐桂琼[1] 

机构地区：[1]上海大学管理学院,上海200444

出　　处：《微电子学与计算机》2017年第7期92-95,100,共5页Microelectronics & Computer

基　　金：国家自然科学基金(11201290)

摘　　要：在多属性模糊c均值聚类的基础上,提出了一种基于属性重要性加权的聚类算法.为验证新算法的有效性,在6个UCI数据集上,将新算法与结合主成分分析法和基于粗糙集指数加权的聚类方法进行了比较分析.实验结果表明,基于属性重要性的聚类算法具有更好的泛用性和稳定性,且随着平均类间簇心距离的增大而提升聚类有效性.Considering the difference of feature attributes, a new clustering algorithm based on the importance of attributes is proposed according to multivariable fuzzy c-means clustering method. In order to verify the validity of new method, the proposed algorithm is analyzed and compared with principal component comprehensive analysis and attributes weighted by rough set theory on six UCI datasets. The experimental results showed that this method has wider scope of application and stability. Moreover, the clustering efficiency increase when datasets have further average distance among cluster centers.

关 键 词：数据挖掘 模糊C均值聚类 属性重要性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]