Overview

Each guide shows how to do one task. Before you use a guide, learn the basic loop in Getting started.

Stream your own data

Feed CSV files, DataFrames, message queues and benchmark datasets to a detector.

Choose a detector

Find candidate detectors for your data and compare them.

Turn scores into alerts

Calibrate scores, choose a threshold and keep anomalies out of the model.

Evaluate a detector

Choose a metric, follow performance over time and report stable results.

Build a pipeline

Add preprocessing, projections, postprocessing and ensembles.

Detect anomalies in seasonal series

Find anomalies in metrics with a daily or weekly cycle.

Run a PyOD model on a stream

Run a PyOD batch detector on a stream.

Run a detector in production

Start from historical data, save the state, run one detector per source and control the time per point.

Write your own detector

Write a detector of your own and add it to PySAD.

Examples

Complete scripts that run end to end.