摘要
针对模糊聚类算法中的噪声敏感性以及点对类的隶属度缺乏典型性的问题,提出一个自适应模糊聚类方法.该方法可以自动地标识那些有影响力的或者说重要的原型样本,反映出这些原型样本对其他样本的影响.又可以自动地标识那些有影响力的或者说重要的类,反映出那些重要的类对其他类的影响.该方法能够有效地降低噪声对有用信息的干扰,为传统的聚类方法提供了一个具有可操作性又有效率的替代方案.该方法的收敛性被理论证明,两个试验检验了它的计算花费和准确性.
Fuzzy c-means (FCM) algorithm has noise-sensitivity and lacks the typicality of membership of point to cluster. An adaptive approach to fuzzy clustering was presented to overcome these problems. The approach could automatically distinguish these important points by multiplying a suitable weight number and these important clusters in dataset. Also, the approach can work well in noisy circumstances and its convergence is proven theoretically. Two experiments were used to verify the effectiveness and efficiency of the proposed approach in this paper.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2004年第10期1280-1284,共5页
Journal of Zhejiang University：Engineering Science
基金
国家"863"高技术研究发展计划资助项目(2002AA41201012)
国家"973"重点基础研究发展计划资助项目(2002CB312200).
关键词
自适应性聚类
隶属度
聚类原型
Adaptive algorithms
Convergence of numerical methods
Fuzzy sets
Theorem proving
