An Improved Method for K_Medoids Algorithm

In this paper, we mainly discuss about k_means and k_medoids algorithm and debate the good properties and shortcomings of the both algorithms, then propose the improving measures for k_medoids algorithm. The main idea is that the method which generates centres of k_medoids algorithm replaced by the way which generates centres of k_means. The computational cost of the improved algorithm is a compromise between k_means and k_medoids. Finding the ´noise´ data in the objects data by examining the distance value vector is another point of the improved algorithm. We examine the improved k_medoids algorithm´s performance in the relevant experiment, and draw the conclusion.