starling-lab / PI_GBM

Code files

• main.py: initial code file to learn model on one fold of data.
• find_best_parameters.py: search best hyperparameters setting of alpha, thredsholds for models.
• eval_fairness_metrics_rate.py: compute scores of fairness metrics of different models.
• eval_performance.py: compute performance scores of learned models on different metrics.

Execute code file by

python find_best_parameters.py or python eval_fairness_metrics_rate.py

Dependency

• Python 3.8.13
• NumPy 1.23.3
• pandas 1.4.4
• scikit-learn 1.1.2
• SciPy 1.9.2
• XGBoost 1.6.2

Data

Due to the size limit, 10 folds cross-validation data can be downloaded from https://www.dropbox.com/s/rj4g5r4nt816guq/data.zip?dl=0