摘要
Fuzzy C-Means(FCM)模糊c均值聚类算法是一个应用广泛、有效的无监督聚类算法。但传统FCM算法存在对所有样本等划分的缺点,导致聚类精度不高、鲁棒性不强。针对上述问题,从整体上引入点密度关系,从局部上引入点邻域信息,用以标记每个样本点,提出基于点密度和邻域信息的模糊c均值算法(DLFCM)。该算法能标记每个不同的样本,克服了FCM算法等划分的缺点,提高了算法的聚类精度和鲁棒性。人造数据集和UCI真实数据集实验验证了该算法的有效性。
Fuzzy c-means(FCM)cluster algorithm is a widely used and an effective unsupervised cluster algorithm.But the traditional FCM algorithm classifies all the samples equally,which leads to low clustering accuracy and low robustness.To deal with this problem,this paper introduces dot density relation from the overall and the local information from the local,to tag every sample point then based on the point density and local information a new FCM algorithm is proposed named DLFCM and proposes a fuzzy C-means algorithm based on dot density and local information(DLFCM).The algorithm can mark each different sample,overcome the disadvantages of FCM algorithm and improve the clustering accuracy and robustness Synthetic data and UCI data experiments have proved the effectiveness of the new algorithm.
作者
吴鹏
WU Peng(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and technology,Shanghai 200093,China)
出处
《软件导刊》
2018年第4期85-88,共4页
Software Guide
关键词
聚类算法
目标函数
邻域信息
鲁棒性
clustering algorithm
objective function
local Information
robustness
