一种改进的K均值聚类算法

An Improved K-Means Clustering Algorithm

针对传统K均值算法中采取的欧氏距离计算相似性的不足,提出一种新的相似性计算方法,并将这种方法与欧氏距离的度量方法进行了比较。在UC I基准数据集上的实验表明,该方法有更稳定的聚类结果,是一种比较有效的聚类度量方法。

According to the disadvantages of calculating similarity based on traditional Euclidean distance of K-Means algorithm,a new similarity metrics method is presented.The given method is compared with the Euclidean distance of the K-Means clustering algorithm.The experiments based on UCI benchmark data sets showed that the method has more stable clustering results,and is an effective clustering metrics method.