基于直觉模糊集的模糊C均值聚类改进算法

Improved fuzzy C-means clustering algorithm based on intuitionistic fuzzy sets

针对特征权重未知且具有直觉模糊数的特征信息的聚类分析问题,提出一种改进的基于直觉模糊集的模糊C均值聚类算法.首先,定义区域密度参数,选择高密度区域中相距最远的样本为初始聚类中心;然后,利用直觉模糊熵计算聚类样本的特征权重,对样本特征值进行加权处理.给出改进的FCM聚类算法的具体步骤,并进行了算例验证.研究结果表明,该算法不仅克服了FCM算法易陷入局部极小值的问题,同时大大减少迭代次数,加快了收敛速度,提高了聚类性能.

An improved fuzzy C-means(FCM)clustering algorithm based on intuitionistic fuzzy sets is proposed for a set of multi-feature clustering analysis problems,in which feature weights are unknown and feature values are intuitionistic fuzzy numbers.A regional density parameter is defined,and c samples with the farthest Euclidean distance from the high-density region are selected as initial clustering centers.The Characteristic values are weighted with Characteristic weights calculated by using intuitionistic fuzzy entropy.Calculation steps for the improved FCM clustering algorithm based on intuitionistic fuzzy sets are given.Validity of the proposed improved algorithm is checked with a numerical example.The improved algorithm can solve the problem of falling into local minima,and greatly reduce iterative time so as to accelerate convergence.