Recall is a classification metric that measures how well a machine learning model identifies actual positive cases. It tells us how many of the positive cases in the dataset were correctly detected by the model.
In simple terms, recall measures how good a model is at finding all the relevant positive cases.
How Is Recall Calculated?
The formula for recall is:
Recall = True Positives รท (True Positives + False Negatives)
A high recall score means the model misses fewer actual positive cases.
Recall vs. Precision and Accuracy
- Recall: Measures how many actual positive cases were correctly identified.
- Precision: Measures how many predicted positive cases were actually correct.
- Accuracy: Measures the percentage of all predictions that were correct.
The choice of metric depends on the type of problem and the cost of different prediction errors.
When Is High Recall Important?
High recall is especially important when missing a positive case can have serious consequences, such as:
- Disease detection
- Fraud detection
- Cybersecurity threats
- Defect detection
- Emergency monitoring
For example, in medical screening, a model with high recall is useful because it reduces the chance of missing a patient who may have a disease.
Benefits of Recall
Some key benefits include:
- Reduces false negatives
- Helps identify more positive cases
- Useful for safety-critical applications
- Provides better insight than accuracy for some imbalanced datasets
Conclusion
Recall is an important machine learning metric that measures a model's ability to identify actual positive cases. It is particularly valuable when false negatives are costly or dangerous. By comparing recall with precision and accuracy, organizations can better understand whether a classification model meets their specific requirements.