空军工程大学学报（自然科学版）2011，Vol.12Issue(1)：78-82,5.DOI:10.3969/j.issn.1009-3516.2011.01.017
无监督多尺度模糊聚类算法研究
Research on Unsupervised Multi -scale Fuzzy Clustering Algorithm (UMFA)
摘要
Abstract
In allusion to the insufficiencies in fuzzy c - means clustering algorithm, this paper advances a new mean shift clustering algorithm. This algorithm called unsupervised multi - scale fuzzy clustering algorithm (UMFA).This algorithm is not influenced by the initializations of the cluster centers, through which the hard clustering partition can be obtained by the use of fuzzy clustering without assuming the data unit and the data center of clustering.By using this algorithm the clustering structure of data can be indicated from different ＂ partition scales＂; and the unsupervised clustering for the given data set can be executed. In order to meet the needs of dealing with the large data sets, this paper proposes a fast unsupervised multi -scale fuzzy clustering algorithm (FUMFA). The experiments prove that UMFA behaves itseff all right on multitudinous data sets, and possesses a best collectivity clustering capability, and the use of it can succeed in indicating the clustering structure of data. Simultaneously the experiments also prove that FUMFA possesses a faster speed and an exact identifiable precision, and is applicable to the large data sets. The experimental results of the two algorithms are satisfactory.关键词
均值漂移/模糊聚类/无监督多尺度聚类Key words
MS/ fuzzy clustering/ unsupervised multi - scale fuzzy clustering分类
信息技术与安全科学引用本文复制引用
魏娜,王建勋,兰文祥..无监督多尺度模糊聚类算法研究[J].空军工程大学学报（自然科学版）,2011,12(1):78-82,5.基金项目
国防科技重点实验室基金资助项目(9140c610301080c6106) （9140c610301080c6106）
