DataTerminology.comNEW
Also known as: Weighted average
A global metric computed by aggregating decisions across all classes, weighting by class support.
DESCRIPTION
Micro-averaging calculates metrics by counting total true positives, false negatives, and false positives across classes. It reflects overall performance weighted by class frequency, favoring majority classes. Commonly used for precision, recall, and F1 in multiclass and multilabel tasks. Choose micro-average when operational throughput and aggregate performance matter. It may mask poor performance on rare classes compared to macro-average.
ALTERNATIVE DEFINITIONS
scikit-learn
If ‘average’ is ‘micro’, calculate metrics globally by counting the total true positives, false negatives and false positives.
Defines global aggregation over all decisions.
Wikipedia
Micro-averaging calculates metrics globally by aggregating contributions of all classes to compute the average metric.
Highlights aggregation across classes and support weighting.
RELATED TERMS
Precision Recall F1 Score Class Imbalance Multiclass Classification Confusion Matrix ROC AUC Averaging Methods Support
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 - Classification metrics and averaging
- Hugging Face - Precision metric with micro averaging
TOP QUESTIONS
When should Micro-Average be used?
Use micro-average to measure overall system performance weighted by class frequency, reflecting aggregate operational impact.
Category: Requirements
How does Micro-Average differ from Macro-Average?
Micro-average aggregates global counts, weighting by support; macro-average treats each class equally regardless of frequency.
Category: Misunderstanding
What is a risk of using Micro-Average?
It can hide poor performance on rare classes, leading to overlooked fairness or compliance issues.
Category: Risk
How is Micro-Average computed programmatically?
Aggregate TP, FP, and FN across all classes, then compute the metric from these totals; most libraries support average=‘micro’.
Category: Management
Which metrics commonly support Micro-Average?
Precision, recall, and F1 frequently support micro-averaging in multiclass and multilabel settings.
Category: Overview
Is Micro-Average good for imbalanced data?
It reflects overall performance but may overweight majority classes; combine with macro-average for balanced insights.
Category: Opinion
What competes conceptually with Micro-Average?
Macro-average emphasizes minority classes by equal weighting, complementing micro-average’s operational perspective.
Category: Alternatives
What is a hot topic around Micro-Average?
Choosing averaging schemes under strict fairness and compliance needs is actively debated in regulated domains.
Category: Topic
Where to learn more about Micro-Average?
Consult scikit-learn’s metrics documentation and the Wikipedia page on precision and recall.
Category: Overview