Welcome to the ACFX documentation

ACFX is Actionable Counterfactual eXplainer. It is a model-agnostic XAI algorithm that provides actionable counterfactual explanations for any type of machine-learning model.

Install

acfx package can be installed from either PyPI or directly from source code GitHub

To install form PyPI:

pip install acfx

To install from source code:

git clone https://github.com/sbobek/acfx
cd acfx/src
pip install .

Examples

• Basic Usage examples
  • Initializing ACFX
  • Prepare the data
  • Prepare bounds
  • Prepare adjacency matrix and causal order
  • (Alternatively) prepare adjacency matrix using external tools
  • Fit to initialize the model
  • Generate Counterfactuals
  • Using custom blackbox
  • Example custom optimizer

Reference

• API reference
  • Explanation interface
  • Counterfactual generation
  • Abstract

Development

• Release notes
• Contributing guide