紧邻类与小类数据集下的模糊聚类有效性指标  

Fuzzy cluster validity index under datasets with adjacent class and small class

作　　者：耿嘉艺 钱雪忠[1] 周世兵[1] Geng Jiayi;Qian Xuezhong;Zhou Shibing(School of Internet of Things Engineering,Jiangnan University,Wuxi Jiangsu 214122,China)

机构地区：[1]江南大学物联网工程学院,江苏无锡214122

出　　处：《计算机应用研究》2020年第9期2651-2655,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61673193);中央高校基本科研业务费专项资金(JUSRP11235)。

摘　　要：模糊聚类有效性指标主要是为了解决模糊C-均值算法需要事先给定最佳聚类数的缺陷,但是现有的大多数模糊聚类有效性指标一般过于依赖聚类质心,使得这类指标在含有紧邻类与大小、密度差异大的数据集上无法准确地判断最佳聚类数。为了缓解这个问题,提出了新聚类有效性指标WS。WS指标在一定程度上考虑了最大最小隶属度法则与模糊集偏差,从而全面展示了数据集的整体信息。在人工与真实数据集上,评估WS指标与现有一些指标的有效性,新指标展现出了较高的准确性。在不同的模糊度下,WS指标表现出了较好的鲁棒性。The fuzzy cluster validity index is mainly to solve the defect that the fuzzy C-means algorithm needs to give the optimal number of clusters in advance,but most of the existing fuzzy cluster validity index are generally too dependent on the cluster centroid,which make it impossible to accurately judge optimal to clustering number in the datasets containing adjacent classes and large differences in size and density. In order to alleviate this problem,this paper proposed a new cluster validity index WS. WS index considered the maximum and minimum membership degree rule and the fuzzy deviation of the dataset to a certain extent,comprehensively showed the overall information of the datasets. In the artificial and actual datasets,this paper evaluated the effectiveness of WS index and some existing indexes. The new WS index shows high accuracy and better robustness.

关 键 词：模糊C-均值 聚类有效性 最佳聚类数 模糊度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]