论文部分内容阅读
利用一种新的距离测度将Dave的广义噪声聚类(GNC)扩展成非欧氏距离的广义噪声聚类(NGNC).模糊C-均值聚类(FCM)和广义噪声聚类都是基于欧氏距离的模型,与它们不同之处在于NGNC是基于非欧氏距离的模型,建立在鲁棒统计观点和势函数基础上,这种非欧氏距离比欧氏距离更加鲁棒,因此NGNC算法比GNC算法更加鲁棒.并且,建立在新的距离测度上的NGNC在处理噪声和野值方面比GNC和FCM更好.实验结果表明了NGNC的良好特性.
A new distance measure is used to extend Dave’s generalized noise clustering (GNC) to non-Euclidean distance generalized noise clustering (NGNC). Fuzzy FCM and generalized noise clustering are both based on Euclidean distance The difference between the models is that the NGNC is based on a non-Euclidean distance model. Based on the robust statistical and potential functions, the non-Euclidean distance is more robust than the Euclidean distance. Therefore, the NGNC algorithm Is more robust than the GNC algorithm, and the NGNC based on the new distance measure is better than GNC and FCM in terms of noise and outliers.The experimental results show the good characteristics of NGNC.