K-Center Clustering Using R

Question

I cannot find a simple library function for clustering k-centers using R, while I could use k-mean ( kmeans() ) and hierarchical clustering ( hclust() ).

Is there a library function for simple greedy clustering of k-centers using R, as shown in this post

If not - since I'm new to R - how to implement it (I understand the logic - just not how to write it in R-code).

Answer 1

Try kmeans with method = "centers" .

If this is not what you are looking for, then CRAN has a Task View cluster with dozens of packages at http://cran.r-project.org/web/views/Cluster.html .

Answer 2

From what is described in this blog post, it looks like one of many sowing strategies used for k-tools. In fact, I would not call this the clustering method, but was pre-clustering or something like that.

Perhaps you should take a look at the flexclust R package, I believe that it has some options and initializations of k-means, and perhaps this option has one option to initialize. Or it could be at http://cran.r-project.org/web/views/Cluster.html

Please note that choosing the facility that is farthest away tends to choose emissions as cluster centers! Take a look, for example. k-means ++, based on a similar idea, but somewhat smarter (plus, it supports randomization better, so you can try several different initializations). Or you can select the object that is closest to (2k-1)/(2k) quantile, which is probably the best guess for a good cluster center.