FCM初始聚类中心敏感性分析与优化
首发时间:2017-06-23
摘要:既有研究认为模糊C-均值聚类对初始聚类中心敏感。算法敏感性受到分类数、数据特征、和模糊指数等因素的影响。给出算法敏感性定义。选取不同特征的数据集,分析算法对初始聚类中心的敏感性。实验验证不同分类情形下算法对初始聚类中心的敏感性和不同初始聚类中心到达给定聚类距离时的收敛速度。实验表明,FCM算法对初始聚类中心并非绝对敏感。对于同一数据集,算法敏感程度和分类数相关。针对算法敏感情形,结合k-means、初始聚类中心优化的最大距离积法,提出模糊指数中心移动法,并分析优化算法的有效性。
For information in English, please click here
Sensitivity Analysis and Optimization of Initial Cluster Centers of FCM
Abstract:Some researches show that fuzzy C- means clustering is sensitive to initial cluster centers. The sensitivity of the algorithm is affected by the clustering number, the data characteristics and the fuzzy exponent. The definition of sensitivity of the algorithm is provided. The sensitivity of the algorithm to initial clusters centers is analyzed by data sets of different characteristics. Some experiments are used for analyze the sensitivity of the algorithm to the initial clustering centers under different clustering cases and the convergence speed of different initial clustering centers reach the specific clustering distance. The results show that the algorithm is not absolutely sensitive to the initial clustering centers, the sensitivity is related to the number of clustering for the same data set. The cluster centers move with fuzzy exponent algorithm is proposed by combining the maximum distances product of k-means to solve algorithm sensitive problem, and the validity of the optimization algorithm is analyzed.
Keywords: fuzzy C- means clustering sensitivity analysis cluster centers move with fuzzy exponent
论文图表:
引用
No.4737704120583514****
同行评议
共计0人参与
勘误表
FCM初始聚类中心敏感性分析与优化
评论
全部评论0/1000