Recall
Of the actual positive cases, how many did the model catch?—measuring false negative rate.
Think of it like
Like a cancer screening test: of the people who actually have cancer, how many does it catch?
Example
There are 100 fraudulent transactions in the data. The detector catches 90. Recall is 90%.
How it actually works
High recall means the model doesn't miss cases. Low recall means it's conservative. Recall alone is incomplete—a model that predicts 'positive' for everything has perfect recall but creates noise. The precision-recall tradeoff is fundamental.
For product teams
For safety-critical tasks (cancer detection, abuse detection), prioritize recall to avoid missing cases.
For engineers
Track recall at different decision thresholds; visualize precision-recall curves.
Related
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome