Prompt
I'm building a fraud detection model and need a dataset with recent transactions and labels
Latest observation
Oct 16, 2026
OpenAI APIWeb search: off
If you need a recent fraud-transaction dataset with labels, the best options are usually a mix of public benchmark datasets and synthetic/commercial data, because truly recent labeled fraud data is often proprietary for privacy reasons.
Good public options
-
IEEE-CIS Fraud Detection
- Large, realistic e-commerce fraud dataset
- Labeled transactions
- Good for feature engineering and tabular models
- Not “recent” in a real-world sense, but widely used
-
Kaggle Credit Card Fraud Detection
- Small, classic benchmark
- Labeled
- Highly imbalanced
- Not recent and anonymized
-
PaySim
- Synthetic mobile money transaction dataset
- Labeled
- Useful if you want transaction sequences and fraud patterns
- Synthetic, not real recent data
-
Synthetic fraud datasets
- You can generate your own with tools like:
sdv/ctgan- custom simulation based on realistic fraud rules
- Best if you need up-to-date schema without privacy issues
- You can generate your own with tools like:
If you need truly recent labeled data
You’ll usually need one of these:
- Your own internal transaction data
- A data vendor specializing in fraud/AML datasets
- A partnership with a bank/payment processor
- A synthetic dataset built to match current transaction patterns
Important note
If “recent” means current month/quarter, public datasets generally won’t satisfy that. Fraud labels also usually lag by days to months, so recency and labels often conflict.
Recommended path
- For prototyping: IEEE-CIS or PaySim
- For production-like realism: generate synthetic data based on your transaction schema
- For real recency: source internal/vendor data
If you want, I can help you with one of these next:
- a list of download links and schema details
- a synthetic fraud dataset generation script
- a feature engineering pipeline for fraud detection