Supplementary Material for Paper "On the Computation of Contrastive Explanations for
Boosted Regression Trees"
ECAI 2024
Contents
This archive contains several folders and files:
- btex contains the code used to construct contrastive reason.
- btex/src/exportmodels/datasets-converted: the original datasets used in the paper.
- btex/src/exportmodels/main.py: the code used to construct models and save them (using PyXAI).
- btex/src/exportmodels/models: trees, theory and instances for each dataset and each model (10 models per dataset).
- XP-BTex: logs of the experiments.
- XP-BTex/extract_data.ipynb: juptyer notebook taht extract data from logs.
Required software
- PyXAI is required to create and save additional models
- Pip package Gurobi (version 10.0.1) and the software gurobi installed on your computer are required in order to compute contrastive explanations.