Confusion matrix

Predicted against actual class, with the count in each cell. Percentages are given per actual class, which is the direction that shows sensitivity and specificity; overall percentages would hide a class the model misses.

Usage

plot_confusion(
  predicted = NULL,
  actual = NULL,
  predicted_labels = NULL,
  actual_labels = NULL,
  positive = NULL,
  show = c("count", "percent", "both"),
  digits = 1,
  title = NULL,
  font = "Arial"
)

Arguments

predicted, predicted_labels

Predicted class labels.

actual, actual_labels

Actual class labels. actual may also be a table or matrix of counts, in which case the labels are taken from its dimnames.

positive

Value of the outcome that counts as the positive class, used to order and label the axes. With positive supplied, the two classes are shown as "Negative" and "Positive"; pass predicted_labels and actual_labels to name them differently.

show

Which numbers to print: "count", "percent" (by actual class), or "both".

digits

Significant digits for percentages.

title

Plot title.

font

Font family used in the plot.

Value

A ggplot object.

Examples

set.seed(1)
actual <- rbinom(200, 1, 0.4)
predicted <- ifelse(runif(200) < 0.8, actual, 1 - actual)
p <- plot_confusion(predicted, actual, positive = 1)