optimalK

Find optimal k of k-Medoid partitions using silhouette widths

optimal.k(x, nk = 10, plot = TRUE, cluster = TRUE, clara = FALSE, ...)

Arguments

x	Numeric dataframe, matrix or vector
nk	Number of clusters to test (2:nk)
plot	Plot cluster silhouettes(TRUE/FALSE)
cluster	Create cluster object with optimal k
clara	Use clara model for large data
...	Additional arguments passed to clara

Value

Object of class clust "pam" or "clara"

References

Theodoridis, S. & K. Koutroumbas(2006) Pattern Recognition 3rd ed.

See also

pam for details on Partitioning Around Medoids (PAM)

clara for details on Clustering Large Applications (clara)

Author

Jeffrey S. Evans <jeffrey_evans<at>tnc.org>

Examples

require(cluster)
#> Loading required package: cluster
#> 
#> Attaching package: 'cluster'
#> The following object is masked from 'package:spatstat.geom':
#> 
#>     volume
  x <- rbind(cbind(rnorm(10,0,0.5), rnorm(10,0,0.5)),
             cbind(rnorm(15,5,0.5), rnorm(15,5,0.5)))

  clust <- optimal.k(x, 20, plot=TRUE, cluster=TRUE)
#> Optimal-K: 2
    plot(silhouette(clust), col = c('red', 'green'))
      plot(clust, which.plots=1, main='K-Medoid fit')

# Extract multivariate and univariate mediods (class centers)
  clust$medoids
#>            [,1]       [,2]
#> [1,] -0.3863414 -0.4797485
#> [2,]  4.8249686  4.9026229
    pam(x[,1], 1)$medoids  
#>          [,1]
#> [1,] 4.692233

# join clusters to data
  x <- data.frame(x, k=clust$clustering)