Recall (Sensitivity)
Recall or Sensitivity (True Positive Rate) · Also known as: Sensitivity, True Positive Rate, TPR
Recall measures the proportion of actual positive cases that were correctly identified by the classifier. It answers the question: 'Of all the cases that were truly positive, how many did we find?' Recall is critical in scenarios where missing positive cases is costly.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use recall when missing positive cases is costly, such as disease diagnosis (missing a cancer diagnosis can be life-threatening), criminal investigation (missing a guilty party), or network intrusion detection (missing a security breach). Recall is less important when false positives are the primary concern, such as in spam filtering.
Strengths & limitations
- Directly measures the ability to identify all positive cases; high recall means few cases are missed.
- Interpretable: recall = 1.0 means no positive cases were missed.
- Essential for applications where false negatives have high cost or safety impact.
- Complements precision to give a complete picture of classification performance.
- Ignores false positives; a model could have high recall while making many incorrect positive predictions.
- Can be artificially high if the model predicts everything as positive.
- Does not account for class imbalance or the relative costs of different error types.
- Alone, recall is incomplete without considering precision.
Frequently asked
When should I prioritize recall over precision?
Prioritize recall when the cost of a false negative is high. In medical diagnosis for serious diseases, missing a diagnosis is dangerous. In security threat detection, missing a threat allows damage. In criminal investigation, missing a guilty party prevents justice. Conversely, in spam filtering, false positives (filtering good email) may be more costly than false negatives (missing some spam).
Can a model have high recall but low precision?
Yes. A model could predict positive for almost everything, correctly identifying most actual positive cases (high recall) but making many false positive errors (low precision). This strategy is useful only when false negatives are extremely costly and false positives are tolerable.
How do recall and precision trade off against each other?
Recall and precision are generally inversely related. Lowering the classification threshold increases recall but decreases precision (more false positives). Raising the threshold decreases recall but increases precision (fewer false positives). Choosing the right threshold involves balancing these two metrics based on the costs of different error types.
Sources
- Fawcett, T. (2006). An introduction to ROC analysis. Pattern Recognition Letters, 27(8), 861-874. DOI: 10.1016/j.patrec.2005.10.010 ↗
- Powers, D. M. (2011). Evaluation: From Precision, Recall and F-Measure to ROC, Informedness, Markedness and Correlation. Journal of Machine Learning Technologies, 2(1), 37-63. link ↗
How to cite this page
ScholarGate. (2026, June 3). Recall or Sensitivity (True Positive Rate). ScholarGate. https://scholargate.app/en/model-evaluation/recall
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Balanced AccuracyModel Evaluation↔ compare
- F1-ScoreModel Evaluation↔ compare
- Matthews Correlation CoefficientModel Evaluation↔ compare
- PrecisionModel Evaluation↔ compare
- SpecificityModel Evaluation↔ compare