区间型数据的可能性聚类算法  

A POSSIBILISITIC CLUSTERING ALGORITHM FOR INTERVAL DATA

作　　者：李庆 贺一民 罗建禄 徐磊 

机构地区：[1]武警警官学院电子技术系,四川成都610213

出　　处：《计算机应用与软件》2013年第10期251-254,共4页Computer Applications and Software

摘　　要：针对区间型数据的模糊c均值聚类(IFCM)算法在实际应用中的不足,将可能性理论引入区间型数据的聚类问题,通过放松样本隶属度的约束条件和修正IFCM算法的目标函数,提出一种区间型数据的可能性聚类算法。通过仿真模拟实验和平均CR指标分析,结果表明:在包含噪声和孤立点等代表性比较差的样本数据的聚类问题中,该算法明显优于IFCM算法,能有效地降低噪声对聚类效果的影响。Fuzzy c means clustering algorithm of interval data has deficiency in practical application. In view of this, we introduce the possibility theory into the clustering problem of interval data, by relaxing the constraints of the sample membership and modifying the objective function of IFCM algorithm, we propose a possibilisitic clustering algorithm for interval data. Through simulation experiments and average CR index analysis, the results show that in the cluster problem containing poorly representative sample data such as noise and outliers, the proposed algorithm is much better than the IFCM algorithm, which can effectively reduce the influence of noise on the clustering result.

关 键 词：区间型数据 模糊C均值聚类 可能性聚类 平均CR指标 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]