Dirty Financial Transactions for Anomaly Detection

Kaggle's Dirty Financial Transactions Dataset is designed for anomaly and fraud detection practice. The dataset contains financial transaction records that are intentionally unclean, requiring preprocessing. Its creator and specific size are not documented.

Use Cases

• Detecting anomalous transactions using features like transaction amount and frequency.
• Training classifiers to flag fraudulent activity based on transaction patterns and metadata.
• Practicing data cleaning workflows on dirty financial records containing missing values and inconsistencies.
• Benchmarking anomaly detection algorithms on transaction data with injected noise or errors.

Strengths

• Dataset is specifically designed for the practical task of anomaly detection.
• Focuses on a high-impact domain with real-world relevance for fraud prevention.

Limitations

• Specific row count, column count, and data completeness metrics are unknown.
• Lack of provenance details like source, time range, and geography limits reproducibility.
• The 'dirty' nature implies inherent label noise and potential class imbalance.

Provenance

Source

Kaggle