摘要： 提取区间型数据的特征值，给出适用于区间型数据模糊聚类的FCM算法族(IFCM)。该算法适用于不同特征样本数据的模糊聚类运算，并可对聚类结果进行优化。聚类效果的仿真比较表明，IFCM聚类的平均失真度比基于欧氏距离的FCM聚类算法低6.81%。由于距离定义的合理性，IFCM可以根据区间型数据的不同特点调整特征值的聚类权重，并推广至多维类型数据的模糊聚类。

关键词: 模糊聚类, 区间型数据, 距离

Abstract: An advanced method, Interval Fuzzy c-Means Clustering(IFCM), is proposed based on a new definition of distance between interval data which is based on eigenvalues of interval data. The new method expands handling objects from single value sets to interval sets compared with general FCM algorithm. The simulations included at the end indicate the validity of IFCM of which average distortion is 6.81% lower than a similar interval FCM algorithm. Moreover, IFCM can handle interval data with different weight of eigenvalues according to different requirement. It also can be expanded easily to multidimensional type data for the rationality of the new distance definition.

Key words: fuzzy clustering, interval data, distance

中图分类号: 

• TP301.6