A fraud model must catch as many fraud cases as possible. Which metric matters most?
Why B is the answer
Recall
Recall measures how many real positives were found.
In detail
- Recall measures how many real cases you catch.
- Recall counts real positives found.
Use case
An illustrative scenario, not a real incident.
Missing a fraud case is costly, so the team optimizes for recall, and accepts more false alarms.
See it in different ways
flowchart TD
A{"Which error hurts most?"} -->|"False alarms"| B["Precision"]
A -->|"Missed cases"| C["Recall"]
A -->|"Both, with imbalanced classes"| D["F1 score"]
A -->|"Numeric predictions"| E["RMSE"]mindmap
root((Metrics))
Classification
Recall: catch every case
Precision: avoid false alarms
Regression
RMSE
Text
ROUGEflowchart TD
Q{"What is worse?"}
Q -->|"Missing a real case"| A0["Optimize recall"]
Q -->|"A false alarm"| A1["Optimize precision"]Key point: Recall counts how many real cases were found.
Why the other options are wrong
- A. Mean absolute error
- MAE is for regression.
- C. Training time
- Time does not measure catches.
- D. Root mean squared error
- RMSE is for regression.