Machine learning and diagnostic testing metric · BEGINNER LESSON
Recall
Recall is a classification metric that measures how many of the actual positive cases a model correctly finds. It is calculated as true positives divided by true positives plus false negatives.
What recall is
Recall is a way to score a model or test when there is a positive class, such as patients who truly have a disease or items that truly belong in a target category. It asks: out of all the real positives, how many were identified? In a confusion matrix, true positives are the real positives the model caught, and false negatives are the real positives it missed. Recall uses those two counts only: true positives divided by true positives plus false negatives.
[1] www.mit.edu › ~edchen93 › postsPrecision, Recall, F1 score, and More | The Ultimate Guide to ...
How it works
A confusion matrix is a table that compares truth with prediction. The rows are the actual classes and the columns are the model’s predictions. Recall is the row-based measure for the positive class. A high recall means few missed positives, which is useful when missing a positive case is costly. Source material also notes that recall is called sensitivity or true positive rate, and in signal detection theory it matches hit rate.
[1] www.mit.edu › ~edchen93 › postsPrecision, Recall, F1 score, and More | The Ultimate Guide to ...
Why it matters
Recall is useful when you care about catching positives, such as catching a rare condition or retrieving relevant items. It is different from precision, which asks how many predicted positives are truly positive. The two often trade off: if a model flags more cases, recall can rise while precision falls. The source also notes that recall is a property of the test itself, unlike precision, which depends on the population where the test is used.
[1] www.mit.edu › ~edchen93 › postsPrecision, Recall, F1 score, and More | The Ultimate Guide to ...
Limits and interpretation
Recall does not say anything about false positives by itself, so a model can have high recall and still raise many false alarms. It also does not give the full picture when classes are imbalanced or when different errors have different costs. The source recommends reporting recall together with precision in those situations. Another limitation is that a recall value alone does not tell you how common the positive class was in the population.
[1] www.mit.edu › ~edchen93 › postsPrecision, Recall, F1 score, and More | The Ultimate Guide to ...
SEE THE IDEA
How recall is computed
All cases that truly belong to the positive class.
Positive cases the model correctly identifies.
Positive cases the model misses.
True positives divided by true positives plus false negatives.
Actual positives are split into true positives and false negatives; recall measures the fraction that were found.
Recall summarizes detection of real positives, not the number of false alarms.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.