摘要
在区间值数据的聚类算法中,区间数之间的距离大多仅考虑两区间数的上下界值,其最大缺陷在于所定义的距离不满足视觉合理性。因此,区间值数据的聚类很难用传统的FCM方法。为了解决这个问题,本文引入了一种新的区间数的距离测度,扩展了一种可直接处理特征空间为区间数的聚类问题的FCM聚类算法。通过对比分析表明,该算法更具合理性及有效性。
In the interval-valued data clustering algorithm,the distance between interval numbers mostly consider the upper and lower bounds value,the biggest defect is that the defined distance does not meet the visual rationality.Therefore,it is difficult to use traditional fuzzy C-means（FCM） to the interval-valued data clustering.In order to solve the problem,a new distance measure for interval numbers was introduced,and FCM clustering algorithm was extended to deal directly with the clustering problem of feature space denoted by interval numbers.By comparing with the traditional method,this method is more effective,more accurate,and more accordant to practice.
出处
《化工自动化及仪表》
CAS
北大核心
2010年第8期26-29,共4页
Control and Instruments in Chemical Industry
关键词
区间数
聚类算法
距离测度
FCM
interval numbers
clustering algorithm
distance measure
FCM
