Form and compare candidate hard partitions of SOM units

Form and compare candidate hard partitions of SOM units

Usage

partition_som(
  ensemble,
  k = ensemble$spec$k,
  method = "ward.D2",
  scope = c("analysis", "all"),
  max_pairwise_comparisons = 1000000L
)

Arguments

ensemble

A fitted som_ensemble.

k

Candidate numbers of clusters. Defaults to the ensemble specification.

method

Hierarchical clustering method used on SOM codebooks.

scope

Evidence scope. The default, "analysis", masks every row not used to fit a given ensemble member. "all" explicitly audits the stability of mapped labels for all rows and must not be used as training- partition evidence in assess_defensibility().

max_pairwise_comparisons

Maximum number of pairwise partition comparisons created across all requested k. This explicit budget guards against quadratic growth in the number of ensemble members.

Value

A som_partitions object. Agreement is reported as ARI and AMI, not accuracy. AMI is chance-adjusted and normalized by the arithmetic mean of the two partition entropies.

Examples

data <- simulate_som_scenario("clusters", n = 45, p = 3, seed = 2)
specification <- som_spec(c(3, 2), seeds = 1:2, rlen = 10, k = 2)
ensemble <- fit_som_ensemble(data, specification, keep_models = FALSE)
partition_som(ensemble)
#> <som_partitions>
#>   fitted partitions : 2 
#>   candidate k       : 2 
#>   evidence scope    : analysis 
#>   pairwise contrasts: 1 
#>   agreement metric  : adjusted Rand index (not accuracy)