# Fraud Detection Synthetic Dataset

## Dataset Description

This is a synthetic dataset for fraud detection created for the XNL LLM Task 3 challenge. It contains transaction data with labeled fraud cases.

### Dataset Summary

- Number of transactions: 34767
- Fraud rate: 10.19%
- Generated using the Synthetic Data Generator tool

### Data Fields

- `transaction_id`: Unique identifier for each transaction
- `user_id`: User who made the transaction
- `timestamp`: When the transaction occurred
- `amount`: Transaction amount
- `merchant`: Where the transaction occurred
- `description`: Text description of the transaction
- `transaction_type`: Type of transaction (purchase, subscription, etc.)
- `device`: Device used for the transaction
- `ip_address`: IP address (for online transactions)
- `location`: Geographic location
- `is_fraud`: Target variable - indicates if the transaction is fraudulent (1) or legitimate (0)

## Additional Information

This dataset was generated using a synthetic data generator that creates realistic transaction patterns with embedded fraud signals. The data can be used for training and testing fraud detection models.

## Argilla Integration

This dataset is also available on Argilla as `fraud-detection-transactions` for interactive exploration and labeling.