📏 Macro Recall

Macro Recall#

The unweighted mean of per-class recall values.

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

What it is#

Recall is the share of actual positives the model catches, \(\text{Recall} = TP/(TP+FN)\). For \(K\) classes, compute recall per class (one-vs-rest) and take the unweighted mean:

\[\text{Recall}_{\text{macro}} = \frac{1}{K} \sum_{i=1}^{K} \text{Recall}_i.\]

Every class counts equally regardless of how many samples it has.

A useful equivalence#

In single-label classification, macro recall is exactly balanced accuracy — the average per-class hit rate. That makes it a go-to headline metric for imbalanced problems, because it refuses to let a dominant class inflate the score.

Macro vs micro vs weighted#

Macro — equal weight per class; good when every class is equally important.
Micro — global TP and FN pooled first; dominated by large classes.
Weighted — per-class recall averaged by the number of true samples in each class.

Worked example#

Three classes with Recall(A)=0.90, Recall(B)=0.60, Recall(C)=0.30:

\[\text{Recall}_{\text{macro}} = \frac{0.90 + 0.60 + 0.30}{3} = 0.60.\]

Even though C is rare, it carries the same weight as A.

Pitfalls and edge cases#

Ignores false positives — recall says nothing about precision, so a model that over-predicts a class can still score well; pair it with macro precision or macro F1.
Empty classes — a class with no true samples has undefined recall and must be handled before averaging.

In code#

from sklearn.metrics import recall_score, balanced_accuracy_score

macro = recall_score(y_true, y_pred, average="macro")
# in single-label problems this equals:
bal_acc = balanced_accuracy_score(y_true, y_pred)

Theme: Classification & Averaging Metrics · All terminology

Hint

Mind map — connected ideas

Macro Precision · Macro F1 · Micro Recall · Macro AUROC (Macro-Averaged AUROC)

Hint

More in Classification & Averaging Metrics

Accuracy · AUC (Area Under the Curve) · Average Precision (AP) · Binary Classification · Classification Probability · Discriminatory Power · F1-score · Gini Coefficient · Harmonic Mean · Log Loss (also called Logarithmic Loss or Cross-Entropy Loss) · Macro AUC · Macro AUROC (Macro-Averaged AUROC) · Macro Averaging · Macro F1