Plot clustering results

Produces a scatter plot matrix (pairs plot) of clustered data, colored by cluster assignment. For 2-column data a single scatter plot is produced instead. Noise points (NA cluster) are shown as grey crosses.

Usage

# S3 method for class 'shoal_clustering'
plot(
  x,
  xcol = NULL,
  ycol = NULL,
  pal = shoal_palette(x$n_clusters),
  col = NULL,
  pch = NULL,
  ...
)

Arguments

x

A clustering result object.

xcol, ycol

Optional column name or index to plot on the x/y axis. When both are supplied, a single scatter plot is produced instead of a pairs matrix.

pal

Character vector of colours for clusters, indexed by cluster ID. Defaults to shoal_palette() for the number of clusters found.

col, pch

Optional per-observation colours and plotting characters, recycled to the number of rows. When given they replace the cluster colouring and the noise crosses respectively, and no legend is drawn.

...

Additional arguments passed to pairs() or plot.default(). main, xlab and ylab given here replace the defaults (the algorithm name and the column names).

Value

x, invisibly.

Details

When xcol and ycol are supplied, a single scatter plot of those two variables is produced instead of the full pairs matrix. Columns can be specified by name or integer index.

Colours come from pal, one per cluster, with noise in grey. To colour points by something other than their cluster, pass col directly: it is recycled to one entry per observation and used as is. The same goes for pch. The cluster legend is drawn only when neither is overridden, since it would no longer describe the points.

Examples

res <- shoal_hdbscan(rings, min_cluster_size = 15L, min_samples = 5L)
plot(res)


# Colour by something else entirely, e.g. an outlier score.
plot(res, col = grey(1 - res$outlier_scores), pch = 19)