Skip to contents

Tidy PAM (Partitioning Around Medoids)

Source: R/unsupervised-clustering.R
tidy_pam.Rd

Performs PAM clustering with tidy output

Usage

tidy_pam(data, k, metric = "euclidean", cols = NULL, ...)

Arguments

data

A data frame, tibble, or dist object

k

Number of clusters

metric

Distance metric (default: "euclidean"). Use "gower" for mixed data types.

cols

Columns to include (tidy select). If NULL, uses all columns.

...

Further arguments passed to pam, such as nstart, variant or starting medoids. cluster.only = TRUE and diss are refused, under any abbreviation R would accept: the first leaves no fit to build the result from, and tidy_pam() sets the second itself from data.

Value

A list of class "tidy_pam" containing:

  • clusters: tibble with observation IDs and cluster assignments

  • medoids: tibble with one row per cluster: the medoid's row position in the data (medoid_index, an integer) and, unless data was a dist object, its values

  • silhouette_avg: average silhouette width

  • silhouette_data: the silhouette information pam() returns (its silinfo)

  • model: original pam object

Examples

# PAM with Euclidean distance
pam_result <- tidy_pam(iris, k = 3)

# PAM with Gower distance for mixed data
pam_result <- tidy_pam(mtcars, k = 3, metric = "gower")