Anomaly Detection Robustness

With this package you can generate categorical data that follows the conditional probabilities from a generated probability graph. Afterwards anomalies can be inserted in the data. Finally, different scenarios can be simulated to study the robustness of an anomaly detection model (e.g. Isolation Forest) to different settings, e.g. number of features.

Getting started

Create a python 3 environment with jupyter using conda env create -f environment.yml and install this package using pip install . from the root directory project.

Open the example notebook to play around.