一种基于K均值预处理回溯的PAM算法

An Improved PAM Based On K-means Data Preprocessing Backward Search Algorithm

针对PAM算法在进行聚类时容易陷入死循环的缺陷,引用了回溯法来解决该问题。但是,加入回溯法的PAM算法具有计算量大迭代次数多的缺点,为了在PAM算法迭代过程中,尽量避免使用回溯法,于是进一步,提出了在进行PAM聚类前,采用K-means算法对数据进行预处理,从而获得粗糙中心点,然后找出一组与粗糙中心点最接近的数据作为初始中心点,再进行PAM聚类。从而得到基于K-means预处理回溯法的PAM算法(K-means Data Preprocessing Backward Search PAM,简称KDPBS-PAM)。实验结果表明,KDPBS-PAM算法极大地改善了PAM算法的性能。

In order to deal with the drawback of the PAM algorithm which is easily falling into the endless loops during clustering,a Backward Search Algorithm comes out.However,the defects of huge calculation and iterative number exist in Backward Search PAM Algorithm.In order to avoid using the Backward Search Algorithm during the iteration,furthermore,K-means Data Preprocessing Algorithm was proposed to get the rough center points before PAM clustering.Then,the closest data to the rough center points are found out to be the initial centers for PAM clustering.Then K-means Data Preprocessing Backward Search PAM is obtained.The experimental result shows that the performance of the PAM has been greatly improved by KDPBS-PAM.