【摘要】 提出了一种新的模糊聚类模型(Fuzzy C-means clustering model,FCM),称为自适应模糊聚类(Adaptive FCM,AFCM).和现有的大多数模糊聚类方法不同的是,AFCM考虑了数据集中全体数据的内在关联性,模型中引入了自适应度向量W和自适应指数p.其中,W在迭代过程中是自适应的,p是一个给定参数.W和p共同作用调控聚类过程.AFCM同时输出三组参数:模糊隶属度集U,自适应度向量W,以及聚类原型集V.本文给出了两组数据实验验证AFCM的性能.第1组实验验证AFCM的聚类性能,以FCM为比较对象.实验表明AFCM可以得到更好的聚类质量,而且通过合理选择自适应指数p,AFCM和FCM在时间复杂性上保持同一水平.第2组实验检验了AFCM的离群点挖掘性能,以目前常用的基于密度的LOF为比较对象.实验表明AFCM算法具有极大的计算效率优势,且AFCM得到的离群点是全局的,反映的是离群点和整个数据集的关系,离群点涵盖的信息也更丰富.文章指出,AFCM在挖掘大数据集和实时数据中的离群点应用方面,以及获得高质量的聚类结果的应用方面,特别在聚类的同时需要挖掘离群点的应用方面具有独特的优势.更多还原

【Abstract】 This paper proposes a new kind of fuzzy C-means clustering model (FCM),which is named as adaptive fuzzy clustering (AFCM). Different from most current fuzzy clustering methods,the AFCM considers the internal connectivity of all data points. An adaptive degree vector W and an adaptive exponent p are introduced into the model to jointly influence the clustering process. The AFCM simultaneously outputs three categories of parameters:fuzzy membership degree matrix U,adaptive degree vector W ,and cluster prototype matrix V. Two groups of numerical experiments,Group 1 and Group 2,were executed to evaluate the AFCM. Group 1 demonstrates the clustering performance of the AFCM,with FCM being its counterpart,and the results showed that the AFCM can obtain better clustering quality,meanwhile its time complexity can hold the same level as that of the FCM by choosing the available p. Group 2 checks the ability of the AFCM in mining the outliers,with the density-based LOF being its counterpart and the results showed that the AFCM has considerable advantages in computing efficiency,and that the outliers minded by the AFCM are global,and reflect the relationship between the outliers and the whole data set. It is pointed out that the AFCM possesses the unique advantages when mining the outliers of the large-scale or dynamic data sets,and clustering the data set for better clustering results,especially when it is necessary to simultaneously fulfill both tasks of clustering and mining outliers.更多还原