Why use F1 instead of accuracy?

Accuracy can be misleading with imbalanced classes. F1 specifically measures performance on the positive class, which is usually what you care about in imbalanced problems.

What is a good F1 score?

Context-dependent. 0.8+ is often considered good. Compare to: random baseline, business requirements, and state-of-the-art for similar problems. Higher isn't always necessary if it meets business needs.

When should I NOT use F1?

When precision and recall have very different business importance. F1 treats them equally. If false negatives are 10x more costly than false positives, use a weighted measure or optimise recall directly.

What's the difference between macro and micro F1?

Macro F1: calculate F1 per class, then average (treats all classes equally). Micro F1: aggregate all predictions, then calculate (larger classes dominate). Use macro for balanced class importance.

Book Free Assessment Calculator

F1 Score

The harmonic mean of precision and recall, providing a single metric that balances both. Useful when you need good performance on both false positives and false negatives.

In-Depth Explanation

The F1 score combines precision and recall into a single metric. The harmonic mean punishes extreme values, requiring both metrics to be high for a good F1.

F1 formula: F1 = 2 × (Precision × Recall) / (Precision + Recall)

F1 characteristics:

Range: 0 to 1 (higher is better)
Harmonic mean (not arithmetic)
Low if either precision OR recall is low
Best when both are balanced

When to use F1:

Imbalanced datasets
Both false positives and negatives matter
Need single metric for model comparison

Variants:

Macro F1: Average F1 across classes
Micro F1: Global TP/FP/FN counts
Weighted F1: Class-weighted average

Business Context

F1 is the go-to metric for imbalanced classification problems where accuracy misleads. It forces you to care about both precision and recall.

How Clever Ops Uses This

We typically report F1 scores for Australian business classification projects, especially when dealing with imbalanced outcomes.

Example Use Case

"Comparing fraud detection models: Model A has F1 of 0.82, Model B has 0.76. Model A better balances catching fraud while minimising false alerts."

Frequently Asked Questions

Learn More

Model Selection and Evaluation: Choosing the Right AI Model for Your Use Case

Learn how to select the optimal AI model for your needs by comparing capabilities, costs, and performance. Includes evaluation frameworks, benchmarking strategies, and migration guidance.

Read article

Experience Mapping Fair Work Compliance