I would like to graphically demostrate the behavior of k-means by plotting iterations of the algorithm from a starting value (at (3,5),(6,2),(8,3)) of initial cluster till the cluster centers. Each iteration may correspond to a single plot with centroids and clusters.
Given:
x<-c(3,6,8,1,2,2,6,6,7,7,8,8) y<-c(5,2,3,5,4,6,1,8,3,6,1,7) df<-data.frame(x,y) dfCluster<-kmeans(df,centers=3) # with 3 centroids
I would like to use the first three tuples as my initial cluster and track the movement of the centroids.
Try to use tryCatch to automate the the process of stopping when conversion is reached: I use the iris-data set because there kmeans needs 2 iterations (the (6,3.5)-Point switches)
set.seed(1337) df = iris[,1:2] dfCluster<-kmeans(df,centers=3, iter.max = 1) plot(df[,1], df[,2], col=dfCluster$cluster,pch=19,cex=2, main="iter 1") points(dfCluster$centers,col=1:5,pch=3,cex=3,lwd=3) max_iter = 10 for (i in 2:max_iter){ tryCatch({ dfCluster <- kmeans(df,centers = dfCluster$centers, iter.max = 1) done <- TRUE }, warning=function(w) {done <- FALSE}) plot(df[,1], df[,2], col=dfCluster$cluster,pch=19,cex=2, main=paste("iter",i)) points(dfCluster$centers,col=1:5,pch=3,cex=3,lwd=3) if(done) break }
Result:
If you want to get the coordinates at each iteration-step see here: Getting the coordinates of every observation at each iteration of kmeans in R
