一种高效鲁棒的无监督模糊c均值聚类算法  

An Efficient and Robust Clustering Algorithm for Unsupervised Fuzzy c-Means

作　　者：曲福恒[1] 胡雅婷[2] 马驷良[3] 郭世龙 李恒燕[5] 

机构地区：[1]长春理工大学计算机科学技术学院,长春130022 [2]吉林农业大学信息技术学院,长春130118 [3]吉林大学数学研究所,长春130012 [4]北京农商银行信息技术部,北京100033 [5]华北水利水电学院数学与信息科学学院,郑州450011

出　　处：《吉林大学学报（理学版）》2012年第6期1179-1184,共6页Journal of Jilin University:Science Edition

基　　金：国家自然科学基金(批准号:10926157);国家"十一五"科技支撑计划项目(批准号:2009BAE69B01)

摘　　要：先通过数据约简技术在不损失数据聚类结构的前提下对数据进行精简,利用提出的近似模糊c均值聚类算法对精简后数据进行划分得到初始化中心,再在该中心基础上通过模糊c均值聚类算法结合聚类有效性指标,实现对数据的无监督聚类,改进了无监督模糊c均值聚类算法聚类性能过分依赖初始化中心及大数据集下计算效率不理想的问题.与已有算法的对比实验表明,所提出的算法具有更高的求解精度与计算效率,得到的聚类个数更合理.On the condition of losing less information and retaining less data, the data were refined by the data reduction technique. The proposed approximation algorithm for fuzzy c-means clustering was used to estimate the cluster centers. Combined with validity indexed and estimated centers, FCM can execute unsupervised clustering. The proposed algorithm improved the computational efficiency and performance of the conventional unsupervised fuzzy c-means clustering algorithm. The contrast experimental results with conventional algorithms show that the proposed algorithm has a relatively high precision and efficiency. It can obtain the cluster number more accurately than the conventional algorithm.

关 键 词：模糊C均值 聚类有效性 无监督聚类 数据约简 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]