Toolboxes & Code Repositories

BELLATREX

Local explainability tool for Random (Survival) Forest predictions

software
Description:

Bellatrex is a tool designed to make Random Forest models more interpretable by generating a small set of diverse rules for a given test instance. This allows users to gain insights into the decision-making process behind the model’s predictions, and is compatible with a wide variety of prediction tasks: (multi-label) classification, (multi-target) regression, and survival analysis.

When provided with a test instance, Bellatrex works by following a series of steps:

  1. Pre-selecting a subset of rules relevant to the prediction.

  2. These rules are then transformed into vector representations.

  3. The vectors are then projected into a low-dimensional space.

  4. Clustering is performed, and one rule from each cluster is selected as part of the explanation.

Each selected rule is visually presented to the user through an intuitive interface, which also allows users to explore similar rules that were not selected. Bellatrex’s user-friendly GUI is designed to make the process of understanding complex models straightforward.

For easy access and integration, Bellatrex is packaged as a Python library, streamlining its use in various machine learning workflows.

More Information:
Publications: