Compute classification metrics

Compute multi-class classification metrics from predicted and true labels, including accuracy, macro-F1, purity, NMI, ARI, and rare-class recall.

Usage

classification_metrics_compute(predicted, truth, rare_threshold = 0.05)

Arguments

predicted

Character vector of predicted class labels.

truth

Character vector of true class labels (same length as predicted).

rare_threshold

Proportion threshold for rare-class recall. Classes with support proportion ≤ this value contribute to the rare_recall metric. Default is 0.05.

Value

A list with the following components:

accuracy

Overall accuracy (scalar).

macro_f1

Macro-averaged F1 score (scalar).

purity

Cluster purity (scalar).

nmi

Normalized Mutual Information (scalar).

ari

Adjusted Rand Index (scalar).

rare_recall

Mean recall on rare classes (scalar, or NA if no rare classes).

class_table

A data frame of per-class precision, recall, F1, and support.

Examples

predicted <- c("A", "A", "B", "B", "C")
truth <- c("A", "B", "B", "B", "C")
classification_metrics_compute(predicted, truth)
#> $accuracy
#> [1] 0.8
#> 
#> $macro_f1
#> [1] 0.8222222
#> 
#> $purity
#> [1] 0.8
#> 
#> $nmi
#> [1] 0.6712695
#> 
#> $ari
#> [1] 0.2105263
#> 
#> $rare_recall
#> [1] NA
#> 
#> $class_table
#>   class precision    recall        f1 support
#> 1     A       0.5 1.0000000 0.6666667       1
#> 2     B       1.0 0.6666667 0.8000000       3
#> 3     C       1.0 1.0000000 1.0000000       1
#>