基于PAM和簇阈值的改进K-Means聚类算法

An Improved K-Means Algorithm Based on PAM Algorithm and Cluster Threshold

为了弥补K-Means算法对孤立点数据敏感的缺陷,提高K-Means算法对包含孤立点数据集的聚类效果,在深入研究K-Means算法的基础上,提出了基于PAM和簇阈值的改进K-Means聚类算法。该算法首先对待聚类数据进行抽样,然后利用PAM算法获取样本数据的聚类中心,以样本数据的聚类中心作为KMeans算法的初始聚类中心。在聚类迭代过程中动态计算各簇阈值,利用簇阈值准确地过滤孤立点数据。实验结果表明,本文提出的算法不仅聚类时间短,而且具有较高的聚类准确率。

In order to overcome the weakness of K‐Means algorithm which is sensitive to outliers ,and to improve the quality of K‐Means clustering algorithm ,the paper makes an in -depth study on the traditional K‐Means algorithm and proposes an improved clustering algorithm based on the PAM algo‐rithm and the cluster threshold .T he proposed method first samples the clustering data and then em‐ploys the PAM algorithm to obtain the clustering center of the sample data as the initial center of K‐Means algorithm .By calculating the threshold for each cluster dynamically in the iterative process of clustering ,the outliers can be excluded from the dataset .Experimental results indicate that the pro‐posed algorithm have been shown lower computational cost and higher clustering accuracy .