一种基于K近邻团的聚类算法

摘要

The concept of k-nearest clique is proposed based on k-nearest neighbors and reversed k-nearest neighbors.To construct k-nearest clique,the similarity between objects and combination all the objects is measured which be k-nearest neighbors and reversed k-nearest neighbors pair wise.Because all the objects in the same clique have a high similarity to each other,the right k can be selected to clustering the object set.The experience shows the effectiveness of the algorithm.%在K近邻和逆K近邻理论基础上提出了K近邻团的概念.通过度量对象间的相似度,任意两个元素都互为K近邻和逆K近邻的对象集合构成一个K近邻团.利用同一个K近邻团中的对象彼此都具有较高相似性的特点,选取不同的K值对目标集合进行聚类.通过实验证明了该方法的有效性.