Jose M. Alvarez, Scuola Normale Superiore, University of Pisa; Pisa, Italy; jose.alvarez@di.unipi.it
Kristen M. Scott, KU Leuven, Leuven.AI; Leuven, Belgium; kristen.scott@kuleuven.be
Bettina Berendt, TU Berlin, Weizenbaum Institute, KU Leuven; Berlin, Germany; berendt@tu-berlin.de
Salvatore Ruggieri, University of Pisa; Pisa, Italy; salvatore.ruggieri@unipi.it
Repository for accompanying code to J.M. Alvarez, K.M. Scott, B, Berendt, and S. Ruggieri. Domain Adaptive Decision Trees: Implications for Accuracy and Fairness. FAccT '23: Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency, June 2023, Pages 423–433, DOI: 10.1145/3593013.3594008.
The data used is under the data folder, though it can be downloaded using FolkTables.
The scripts are under the source folder. It is necessary to run the notebook 01_CalculateDistances first. The notebook 02_PlotDistances analysis these results. The notebook 03_TestSingleSourceTarget presents an example on how to implement DADT. The notebook 04_AnalyseResults carries out the analysis presented in the paper and should be run after running script experiments.py.
The results are stored under the results folder.