Skip to contents

Plot anomaly boxplot with feature distributions

Source: R/anomaly_plot.R
plot_anomaly_boxplot.Rd

Create boxplots showing how anomalous points compare to the overall feature distributions. All data points are shown as small black dots, with the anomalous point(s) highlighted in red (or custom color) for easy identification. Can work with or without a contribution object.

Usage

plot_anomaly_boxplot(
  contribution_obj = NULL,
  data,
  sample_id = NULL,
  top_n = 5,
  highlight_color = "red",
  highlight_size = 3,
  highlight_shape = 17,
  show_contribution = TRUE
)

Arguments

contribution_obj

A feature_contribution object from feature_contribution(), or NULL. If NULL, you can specify multiple sample_ids to highlight multiple anomalies.

data

The original training data

sample_id

The sample ID(s) to visualize. Can be a single value or a vector.

  • If contribution_obj is provided: only single value is used

  • If contribution_obj is NULL: can be a vector to mark multiple anomalies

  • If NULL and contribution_obj provided: uses first sample from contribution_obj

top_n

Number of top contributing features to display (default 5, NULL for all). Only used when contribution_obj is provided.

highlight_color

Color for the anomaly point(s) (default "red")

highlight_size

Size of the anomaly point(s) (default 3). Automatically adjusted to 1 when >5 anomalies are highlighted

highlight_shape

Shape of the anomaly point(s) (default 17, not used currently)

show_contribution

Whether to show contribution percentages in labels (default TRUE). Only applies when contribution_obj is provided.

Value

A ggplot object

Examples

# With contribution object
model <- isoForest(iris[1:4])
contributions <- feature_contribution(model, sample_ids = 42, data = iris[1:4])
plot_anomaly_boxplot(contributions, iris[1:4], sample_id = 42)


# Without contribution object - mark multiple anomalies
anomaly_ids <- c(42, 107, 119)
plot_anomaly_boxplot(contribution_obj = NULL, data = iris[1:4], sample_id = anomaly_ids)


# Many anomalies (auto-adjust size)
many_ids <- c(42, 61, 99, 107, 119, 132, 135)
plot_anomaly_boxplot(NULL, iris[1:4], many_ids)
#> Detected 7 anomalies (>5), adjusting point size to 1