Calculate Posterior Fraud Probability Using Bayes' Theorem
Overview
This question evaluates understanding of Bayes' theorem, conditional probability, and probabilistic reasoning in the context of classifier metrics (prior probability, true positive rate, false positive rate) for fraud detection.
Meta
Aug 4, 2025, 10:55 AM
Data Scientist
Onsite
Statistics & Math
11
0
Posterior Fraud Probability After a Flag
Context
You operate a fraud detection system that flags accounts as suspicious. Define:
- F: account is fraudulent
- +: system flags the account
- Prior fraud rate p = P(F)
- True positive rate (TPR) = P(+ | F)
- False positive rate (FPR) = P(+ | not F)
Task
- Use Bayes’ theorem to derive an expression for the posterior probability that an account is fraudulent after a flag: P(F | +).
- Compute P(F | +) for the following example values: p = 1%, TPR = 90%, FPR = 5%.
- Briefly interpret the result for decision thresholds (e.g., how the posterior compares to a 10%, 20%, or 50% action threshold).
Solution
Login required
Show