Self-Organizing Map (SOM) clustering

Trains a Self-Organizing Map on the input data using kohonen::som() and returns the mapping-unit (node) assignment for every observation as a cluster label.

Usage

cluster_som(
  data,
  vars = NULL,
  x_dim = 10L,
  y_dim = 10L,
  rlen = 10L,
  topo = "rectangular"
)

Arguments

data

matrix or data.frame. Rows are observations, columns are variables.

vars

character vector or NULL. Column names to use for clustering. If NULL (default), all columns are used.

x_dim

integer(1). Number of columns in the SOM grid. Default 10L (matching FlowSOM).

y_dim

integer(1). Number of rows in the SOM grid. Default 10L (matching FlowSOM).

rlen

integer(1). Number of training iterations. Default 10L (matching FlowSOM).

topo

character(1). Grid topology passed to kohonen::somgrid(). One of "rectangular" (default) or "hexagonal".

Value

A character vector of cluster labels, one per row of data. Labels are of the form "1", "2", etc., corresponding to SOM grid nodes.

Details

The function:

1. Scales the selected columns to zero mean and unit variance.

2. Creates a SOM grid of size x_dim × y_dim.

3. Trains a SOM on the scaled data.

4. Returns the unit assignment for each observation.

The default grid size (10 × 10) and training length (rlen = 10) match the defaults used by FlowSOM's BuildSOM.

The kohonen package must be installed.

Examples

set.seed(1)
mat <- matrix(
  c(rnorm(50, -3), rnorm(50, 3), rnorm(100, 0)),
  ncol = 2,
  dimnames = list(NULL, c("v1", "v2"))
)
cl <- cluster_som(mat, x_dim = 3, y_dim = 3)
table(cl)
#> cl
#>  1  2  3  4  5  6  7  8  9 
#> 14  3 15 18  3 15 12  1 19