⬡PipelineFinanceFree
PyOD — Outlier & Anomaly Detection
A comprehensive Python toolkit of 60+ outlier-detection algorithms, widely applied to financial fraud and anomaly detection.
PyOD
PyOD is one of the most widely used Python libraries for detecting anomalous data points, offering more than 60 detection algorithms under a single, scikit-learn-style API. While general-purpose, it is a staple for financial fraud and anomaly detection, where flagging unusual transactions, accounts, or market behavior is the core task.
Key features
- 60+ detectors from classical (LOF, Isolation Forest, KNN) to deep learning (AutoEncoder, DeepSVDD)
- Unified
fit/predict/decision_functionAPI compatible with scikit-learn - Model-combination utilities (averaging, maximization) for robust ensembles
- Benchmarks and thresholding helpers for setting anomaly cutoffs
- Extensively documented, tested, and cited across academia and industry
A typical fraud workflow fits a detector on transaction features, then scores new records; high anomaly scores surface likely-fraudulent activity for review.
Curated mirror of the open-source PyOD (BSD-2-Clause). Get it from the source.