DataTerminology.comNEW
Also known as: Unweighted average
An unweighted mean of per-class metric scores in multiclass or multilabel evaluation.
DESCRIPTION
Macro-averaging computes a metric independently for each class and then averages them equally. It treats all classes with equal importance, making it robust to class imbalance when minority classes matter. This averaging is often used for recall, precision, and F1 score in multiclass tasks. It can reveal poor performance on rare classes that micro-averaging may hide. Practitioners choose macro-average for fairness across classes or regulatory reporting.
ALTERNATIVE DEFINITIONS
scikit-learn
If ‘average’ is ‘macro’, calculate metrics for each label, and find their unweighted mean. This does not take label imbalance into account.
Explicitly notes class imbalance is not accounted for.
Wikipedia
Macro-averaging calculates metrics for each label, and finds their unweighted mean; it does not take label imbalance into account.
Standard description used across the ML community.
RELATED TERMS
Precision Recall F1 Score Balanced Accuracy Class Imbalance Multiclass Classification Confusion Matrix ROC AUC Averaging Methods
COMPETING TERMS
KNOWLEDGE AREAS
Data Science AI Testing & Validation
TARGET AUDIENCE
Data Scientist AI Engineer MLOps Engineer Data Analyst
EXTERNAL SOURCES
- Wikipedia - Precision and recall
- scikit-learn - Overview of classification metrics
- Hugging Face - F1 metric with averaging options
TOP QUESTIONS
When should Macro-Average be used?
Use macro-average when minority classes are as important as majority classes. It highlights performance on rare classes by giving equal weight to each class.
Category: Requirements
How does Macro-Average differ from Micro-Average?
Macro-average computes metrics per class and averages them equally, while micro-average aggregates over all decisions, weighting classes by their support.
Category: Misunderstanding
What is a risk of using Macro-Average?
It can underrepresent overall accuracy if many minor classes exist, possibly overemphasizing rare labels compared to practical business priorities.
Category: Risk
What metrics commonly use Macro-Average?
Precision, recall, and F1 score often support macro-average in multiclass and multilabel evaluation settings.
Category: Overview
Is Macro-Average suitable for imbalanced datasets?
Yes, if you value minority classes equally. However, it ignores actual class frequencies and may not reflect aggregate performance.
Category: Opinion
How is Macro-Average computed programmatically?
Compute the metric per class, then take the arithmetic mean across classes. Most libraries provide an average=‘macro’ parameter.
Category: Management
What is the biggest misunderstanding about Macro-Average?
Some think it handles imbalance automatically; it doesn’t. It only equalizes class weights regardless of frequency.
Category: Misunderstanding
What is the competing term for Macro-Average?
Micro-average aggregates decisions across classes, effectively weighting by support, and often reflects overall operational performance.
Category: Alternatives
Where to learn more about Macro-Average?
Refer to scikit-learn’s classification metrics and the Wikipedia page on precision and recall for definitions and examples.
Category: Overview