Optimize the Silhouette Width of K-Means Clustering Solutions

Generates k-means solutions from 2 to nrow(d) - 1 number of clusters and returns the number of clusters with a higher silhouette width median. See utils_cluster_silhouette() for more details.

Usage

utils_cluster_kmeans_optimizer(d = NULL, seed = 1)

Arguments

d

(required, matrix) distance matrix typically resulting from distantia_matrix(), but any other square matrix should work. Default: NULL

seed

(optional, integer) Random seed to be used during the K-means computation. Default: 1

Value

data frame

See also

Other internal_dissimilarity_analysis: utils_block_size(), utils_cluster_hclust_optimizer(), utils_cluster_silhouette(), utils_importance_df_to_wide()

Examples

#parallelization and progress bar
#for large datasets, parallelization accelerates cluster optimization
future::plan(
  future::multisession,
  workers = 2 #set to parallelly::availableWorkers() - 1
)

#progress bar
# progressr::handlers(global = TRUE)

#weekly covid prevalence
#in 10 California counties
#aggregated by month
tsl <- tsl_initialize(
  x = covid_prevalence,
  name_column = "name",
  time_column = "time"
) |>
  tsl_subset(
    names = 1:10
  ) |>
  tsl_aggregate(
    new_time = "months",
    fun = max
  )

if(interactive()){
  #plotting first three time series
  tsl_plot(
    tsl = tsl_subset(
      tsl = tsl,
      names = 1:3
    ),
    guide_columns = 3
  )
}

#compute dissimilarity matrix
psi_matrix <- distantia(
  tsl = tsl,
  lock_step = TRUE
) |>
  distantia_matrix()

#optimize hierarchical clustering
kmeans_optimization <- utils_cluster_kmeans_optimizer(
  d = psi_matrix
)

#best solution in first row
head(kmeans_optimization)
#>   clusters silhouette_mean
#> 1        2       0.3175009
#> 2        5       0.3080265
#> 3        4       0.2970758
#> 4        3       0.2618570
#> 5        6       0.2460827
#> 6        7       0.2312105

#disable parallelization
future::plan(
  future::sequential
)