改进的基于距离阈值的FCM算法  

A fuzzy clustering algorithm based on a distance threshold

作　　者：李玲玲[1] 

机构地区：[1]淮北职业技术学院计算机科学技术系,安徽淮北235000

出　　处：《贵州师范学院学报》2011年第9期11-14,共4页Journal of Guizhou Education University

基　　金：国家自然科学基金(NO.60705015)

摘　　要：凝聚型层次聚类和模糊C-均值聚类是聚类中的两种常用算法,每种算法都有其自己的优点、缺点及适用的对象和范围。针对FCM算法的对初始值敏感,以及目标函数没有考虑类间距离的缺点,通过使用距离阈值,把凝聚型层次聚类与模糊C-均值聚类算法相结合,产生一种新的基于距离阈值的FCM算法,实验结果表明,这种算法能够自动的判断迭代的终止条件、快速有效的找到最佳聚类结果,从而实现对模糊C-均值聚类算法的自动优化。Agglomerative hierarchical clustering and fuzzy c -means （FCM） are the basic methods in clustering. Each of them has its own advantages and disadvantages, and their corresponding scope and objects are very different. Aiming at the sensitivity of initial value in fuzzy c - means （FCM） and the defect of unconsidered distance between classes in objective function, the paper combines agglomerative hierarchical clustering and FCM to produce a new FCM method based on distance threshold. The result of the experiment shows that this method can automatically judge the ending time and can find the best clustering result, so the method can optimize automatic which is better than the FCM.

关 键 词：凝聚型层次聚类 FCM 距离阈值 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]