采用不确定性度量的粗糙模糊C均值聚类参数获取方法  

Selection Method for Parameters of Rough Fuzzy C-Means Clustering Based on Uncertainty Measurement

作　　者：王学恩[1,2] 韩德强[1,2] 韩崇昭[1,2] 

机构地区：[1]西安交通大学智能网络与网络安全教育部重点实验室,西安710049 [2]西安交通大学电子与信息工程学院,西安710049

出　　处：《西安交通大学学报》2013年第6期55-60,共6页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金群体创新基金资助项目(60921003);国家自然科学基金资助项目(61074176;61104214)

摘　　要：针对粗糙模糊C均值聚类的阈值、权重选取问题,提出了一种基于不确定性度量的参数自适应获取方法。该方法将阈值选取归结为一个最优划分寻找问题,给出一种基于方差的划分优劣评价方法;利用信息熵来度量样本归属的模糊性,基于该模糊性度量和类簇的粗糙度,提出了一种权重参数自适应计算方法。将所提方法应用于粗糙模糊C均值聚类,并将分别基于所提方法与典型参数选取方法的粗糙模糊C均值聚类算法在人工数据集和真实数据集上进行实验比较。结果表明,基于所提参数确定方法的粗糙模糊C均值聚类能获得更好的聚类有效性和准确性。A selection method for parameters of the rough fuzzy C-means clustering based on uncertainty measurement is proposed.The selection of parameter threshold is converted into an optimal partition problem,and partitions are evaluated based on variances.The information entropy is employed to measure the fuzziness of the sample membership to clusters,then an adaptive method to calculate weights is presented based on the roughness of clusters and the fuzziness of samples.The adaptive parameter selection method is employed in rough fuzzy C-meaning algorithm.Experiments and comparisons with some existing typical parameters selection methods for rough fuzzy C-means clustering algorithm are performed on synthetic and real-world data sets,and the results show that the proposed algorithm achieves higher accuracy,and is effective.

关 键 词：聚类 粗糙模糊C均值 熵 粗糙度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]