Anomaly Detection Python Library : Arundo-adtk

Arundo-adtk

Site : https://arundo-adtk.readthedocs-hosted.com

Anomaly Detection Toolkit (ADTK) is a Python package for unsupervised / rule-based time series anomaly detection.

As the nature of anomaly varies over different cases, a model may not work universally for all anomaly detection problems. Choosing and combining detection algorithms (detectors), feature engineering methods (transformers), and ensemble methods (aggregators) properly is the key to build an effective anomaly detection model.

This package offers a set of common detectors, transformers and aggregators with unified APIs, as well as pipe classes that connect them together into a model. It also provides some functions to process and visualize time series and anomaly events.

 

Anomaly Types in Arundo-adtk 

Outlier

Spike

 

Level Shift

Pattern Change

Seasonality