作者：刘永利；段天毅；杨立身

第一作者单位：河南理工大学计算机科学与技术学院

摘要：为了同时对数据对象和特征进行聚类分析以提高聚类准确率,在模糊紧致性和分离性算法(fuzzy compactness and separation,FCS)基础上,提出一种结合类内紧致性和类间分离性的模糊联合聚类算法(fuzzy compactness and separation co-clustering,FCSCC)。该算法在FCS的基础上增加了对特征维度的隶属度关系与熵最大化原理,能够在数据对象和特征2个维度上同时聚类。为验证该算法的有效性,另选择了3种算法在5个数据集上进行了对比实验,结果表明,FCSCC算法的聚类准确率高于其他3种算法。

Abstract：In order to implement clustering on feature dimension as well as data object dimension and improve clustering accuracy, a fuzzy compactness and separation co-clustering algorithm (FCSCC) is proposed based on the fuzzy compactness and separation algorithm (FCS) .In FCSCC, feature membership and entropy maximization is added into FCS.It can simultaneously group data objects and features.In order to evaluate clustering effectiveness, experiments were carried out on five datasets to compare the FCSCC with other three clustering algorithms.The experimental results show that the FCSCC algorithm is better than these three methods in terms of accuracy.

基金：国家自然科学基金资助项目(61202286)；河南省高等学校青年骨干教师科研项目(2015GGJS-068)；河南省高校基本科研业务费专项资金资助项目(NSFRF1616)；

关键词：模糊联合聚类算法;紧致性;分离性;

DOI：10.16186/j.cnki.1673-9787.2017.05.014

分类号：TP311.13