pyod

Why we included this project

Anomaly detection shows up in nearly every data pipeline, and pyod is one of the few libraries that treats it as a general problem rather than a single algorithm. It bundles more than sixty detectors, from classical statistical methods to deep-learning models, and they work on tabular, time-series, graph, text, image, and audio data alike. The unified fit/predict API means you can run a whole battery of detectors on the same dataset and keep the ones that actually catch the fraud, intrusions, or manufacturing defects you care about. The newer ADEngine layer automates detector selection and evaluation, and an agentic skill plus MCP server let an AI agent drive the investigation in natural language while the classic API stays intact. If you want a broad, dependable baseline for outlier detection instead of betting on one specialized model, this is a sensible place to start.