Skip to contents

Mirrors the C# GoodnessOfFit classification statics (Accuracy/Precision/Recall/ F1Score/Specificity/BalancedAccuracy): both observed and modeled are already-binary label vectors, compared elementwise – a value equal to its counterpart counts as a match. There is no threshold argument, in R or in C#; threshold your own series into 0/1 labels before calling this.

Usage

classification_metrics(observed, modeled)

Arguments

observed, modeled

numeric vectors of equal length holding binary (0/1) labels.

Value

a named numeric vector of accuracy, precision, recall, f1, specificity, and balanced_accuracy.

Examples

obs <- c(1, 0, 1, 1, 0)
mod <- c(1, 0, 0, 1, 0)
classification_metrics(obs, mod)
#>          accuracy         precision            recall                f1 
#>        80.0000000         1.0000000         0.6666667         0.8000000 
#>       specificity balanced_accuracy 
#>         1.0000000         0.8333333