摘要
为了研究基于马氏距离模糊聚类算法的有效性,首先对比分析了基于数据集模糊划分与几何结构的模糊聚类有效性指标,确定了将紧致度、分离度与清晰度结合的有效性研究方向,然后针对基于马氏距离的模糊聚类提出新的度量标准,构造有效性指标,最后结合算法在真实数据集上进行实验,结果表明新指标能准确识别马氏距离模糊聚类算法在多维数据上的最佳聚类数目。
In order to study the validity of the fuzzy clustering algorithm based on Mahalanobis distance,we have contrasted and analyzed fuzzy clustering validity index based on data set fuzzy partition and geometric structure firstly,and determined the validity research direction of combining firmness and separation with definition.Then the paper proposed new metrics according to fuzzy clustering based on Mahalanobis distance and constructed the validity index.Finally it was tested with algorithm on real data sets.The result shows that the new index can accurately identify the best clustering number when the algorithm makes multi-dimensional data clusterings.
作者
祖志文
李秦
ZU Zhi-wen;LI Qin(College of Mathematics and Physics,Lanzhou Jiaotong University, Lanzhou 730070,China)
出处
《陕西理工大学学报（自然科学版）》
2018年第2期33-38,共6页
Journal of Shaanxi University of Technology:Natural Science Edition
基金
国家自然科学基金资助项目(A020205)
关键词
马氏距离
模糊聚类
有效性指标
最佳聚类数
Mahalanobis distance
fuzzy clustering
validity index
optimal cluster number
