-
Run run_DiCE_experiments_figure1and2.py: This will generate CFs for given data and configuration and stores them in figure1_experiment_results folder. It also runs Figure 2 experiments and stores the results in figure2_experiment_results folder. The code needs to be edited slightly (should add data characteristics and model path) if we need to test with data other than the four we used. For our sample data, I have already run this code and have stored the results.
-
Run get_figure1_summary_stats.py: This will compute the summary statistics required for plotting Figure 1 and store them in figure1_summary_stats folder.
-
Run run_LIME_experiments: This will generate LIME explanations and store them in lime_explanations folder.
-
Run get_figure2_summary_stats.py: This will compute the summary statistics required for plotting Figure 2 and store them in figure2_summary_stats folder. The code needs to be edited slightly if we need to plot Figure 2 for linear models.
-
Run DiCE_evaluation_plotting.ipynb: This notebook generates Figure 1 and 2 plots.
- datasets: Contains COMPAS, Lending Club, and German credit card data. The code downloads Adult income data from the internet.
- paper_plots:: Contains the plots that are in the paper.
- stored_ml_models: Contains the trained ML models of the four datasets.
Please ignore get_mixedIntCF_summary_stats.py and mixedIntCF_results as they are incomplete. These are used in generating Figure 1 plots for linear models.