自动调整样本和特征权值的模糊聚类算法  

Fuzzy clustering algorithm based on the automatic variable weights of samples and features

作　　者：李凯 高岩 曹喆 LI Kai;GAO Yan;CAO Zhe(School of Cyber Security and Computer,Hebei University,Baoding 071002,China)

机构地区：[1]河北大学网络空间安全与计算机学院,河北保定071002

出　　处：《哈尔滨工程大学学报》2018年第9期1554-1560,共7页Journal of Harbin Engineering University

基　　金：国家自然科学基金项目(61375075);河北省自然科学基金项目(F2018201060);河北大学自然科学基金项目(799207217074)

摘　　要：针对模糊c均值聚类算法对特征噪声和样本噪声较敏感的缺陷,依据特征和样本对聚类的不同影响,将特征权值和样本权值引入到模糊c均值聚类的目标函数,并获得了一个模糊聚类模型。利用拉格朗日方法对该模型求解,提出了样本和特征权值自动调整的模糊聚类算法;同时,将核策略引入到该模糊聚类模型,提出了样本和特征权值自动调整的核模糊聚类算法。实验结果表明该方法对含有特征噪声与样本噪声数据的聚类具有较好的处理能力,为特征提取与样本选取等问题提供了一种可行的途径。To improve the sensitivity of the fuzzy c-means clustering algorithm to feature and sample noise,feature and sample weights were introduced to the objective function of the fuzzy c-means clustering algorithm to develop a new fuzzy clustering model.The model was solved using the Lagrange method,and a fuzzy clustering algorithm in which the sample and feature weights could be automatically adjusted was proposed.The kernel strategy was introduced to the new fuzzy clustering model,and a kernel fuzzy clustering algorithm in which the sample and feature weights could be automatically adjusted was proposed.The experimental results showed that the method features excellent processing ability for clustering of data containing feature and sample noise;it provides a feasible approach for feature extraction and sample selection.

关 键 词：模糊聚类 目标函数 样本与特征加权 样本加权 特征加权 核方法 特征噪声 样本噪声 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]