To run this code first install the packages in requirements.txt. This code also requires custom extensions of the open sourced actionable-recourse and LIME libraries contained in actionable-recourse/ and lime/ respectively. Follow the requirements instructions and run python setup.py in each folder.

This file contains the data loading and pre-processing modules.

This file contains the LR, DNN, and SVM implementations.

This file contains the implementations of all the recourse methods. Our method, ROAR, is called RobustRecourse, CFE is called counterfactual_recourse, AR is called actionable_recourse, and CR is called CausalRecourse.

This file contains helper functions for the recourse methods and evaluation.

This file generates recourses for our real world data experiments.

This file generates recourses for our synthetic experiments.

This file generates recourses for our experiments varying $\delta_max$ (see Appendix).