Explain unsupervised fraud and evaluation

Q: Explain unsupervised fraud and evaluation

This is a Machine Learning interview question from PayPal for Data Scientist roles. View the full question and solution on PracHub.

Q: How do I approach Machine Learning interview questions?

Machine Learning questions require understanding of core concepts and practice. PracHub provides solutions with explanations to help you master machine learning interviews.

Question

Unsupervised Fraud Detection: Methods, When to Use Them, and How to Evaluate Without Reliable Labels

Context

You are designing fraud detection for a large payments platform. Fraud is rare and evolving, labels (e.g., chargebacks) are delayed or incomplete, and you have a limited manual review budget. You need to:

Explain when you would use unsupervised approaches versus supervised methods.
Compare common unsupervised options: clustering, density estimation, Isolation Forests, autoencoders, and graph-based anomaly detection.
Describe how to evaluate models without reliable labels, including:
- Precision@k, recall at a fixed review budget, PR-AUC vs ROC-AUC under extreme imbalance, and other rank-based metrics.
- Using proxy/delayed labels and calibration checks.
Clarify why raw accuracy is misleading for this problem and how to choose thresholds under operational constraints.

Explain unsupervised fraud and evaluation

Unsupervised Fraud Detection: Methods, When to Use Them, and How to Evaluate Without Reliable Labels

Context

Solution

Comments (0)