Designing an XAI Interface for Tree-Based ML Models

Supplementary material

Contents

• sources
• sources/pyxai-xai/pyxai/examples/ecai24: this directory contains the code related to the experiments done in the paper (creation of the user, cases, statistics). In particular, the main file (main.py) is the one to be launched.
• dataset: The datasets converted and used in our experiments.
More precisely, for each dataset, you can find:
• datasets/<dataset>.csv: The converted dataset (one-hot encoding of categorical features).
• datasets/<dataset>.types: A JSON file containing all the information about the features.
• logs: The outputs produced by the algorithm run in the experiments.
• times.pdf: Computation times (in seconds) required by the interactions for each dataset used in the experiments.

Software

Setup

• Be sure to use a Linux OS and to use a version of Python >= to 3.8.
• Install Pyxai. Follow these instructions. Instead of cloning the sofware, please use the source provided in this archive.

How to use our program

• For a given dataset, the program sources/pyxai-xai/pyxai/examples/ecai24/main.py implements our protocol.

cd sources/pyxai-xai/pyxai/
export PYTHONPATH=..
python3 examples/ecai24/main.py -dataset=../../../datasets/compas.csv -types=../../../datasets/compas.types


• The program outputs all information in the console.