Skip to contents

Plot a confusion matrix with the full set of derived measures

Source: R/FUNCTIONS_TRAIN_TEST.R
plot_confusion_extended.Rd

Draws a 2x2 confusion matrix (raw counts, shaded blue by magnitude) alongside the row, column and overall measures:

Total measures

Accuracy, Prevalence, Proportion Incorrectly Classified

Horizontal measures

Sensitivity / Miss Rate (observed positive column), Specificity / Fall-out (observed negative column)

Vertical measures

Precision / False Discovery Rate (predicted positive row), Negative Predictive Value / False Omission Rate (predicted negative row)

Usage

plot_confusion_extended(
  observed,
  predicted,
  positive = NULL,
  base_size = 10,
  title = ""
)

Arguments

observed

Vector of true class labels (2 unique values).

predicted

Vector of predicted class labels (2 unique values, same domain).

positive

Value treated as the positive class. Defaults to the second sorted level (matches rwf's confusion() convention).

base_size

Base font size. Default 10.

title

Plot title suffix.

Details

With \(TP\), \(TN\), \(FP\) and \(FN\) the numbers of true positives, true negatives, false positives and false negatives, and \(N = TP + TN + FP + FN\) the total number of cases:

Accuracy

\(\frac{TP + TN}{N}\)

Prevalence

\(\frac{TP + FN}{N}\)

Proportion Incorrectly Classified

\(\frac{FN + FP}{N} = 1 - \mathrm{Accuracy}\)

Sensitivity (true positive rate, recall)

\(\frac{TP}{TP + FN}\)

Miss Rate (false negative rate)

\(\frac{FN}{TP + FN} = 1 - \mathrm{Sensitivity}\)

Specificity (true negative rate)

\(\frac{TN}{FP + TN}\)

Fall-out (false positive rate)

\(\frac{FP}{FP + TN} = 1 - \mathrm{Specificity}\)

Precision (positive predictive value)

\(\frac{TP}{TP + FP}\)

False Discovery Rate

\(\frac{FP}{TP + FP} = 1 - \mathrm{Precision}\)

Negative Predictive Value

\(\frac{TN}{FN + TN}\)

False Omission Rate

\(\frac{FN}{FN + TN} = 1 - \mathrm{NPV}\)

Examples

plot_confusion_extended(observed=c(0,0,0,1,1,1,1),
                        predicted=c(0,0,0,1,1,1,1))

plot_confusion_extended(observed=c(0,0,0,1,1,1,1),
                        predicted=c(0,0,0,1,1,1,0),
                        positive=0)

plot_confusion_extended(observed=c(0,0,0,1,1,1,1),
                        predicted=c(0,0,0,1,1,1,0),
                        positive=1)