Machine learning interpretability for a stress scenario generation in credit scoring based on counterfactuals

To boost the application of machine learning (ML) techniques for credit scoring models, blackbox problem should be addressed. The primary aim this paper is to propose a measure based on counterfactuals evaluate interpretability ML technique. Counterfactuals assist with understanding model regard classification decision boundaries and robustness. second contribution development data perturbation technique generate stress scenario. We apply these two proposals dataset UK unsecured personal loans compare logistic regression stochastic gradient boosting (SBG). show that training (SGB) as conditioned our can provide insight into performance under stressed scenarios. empirical results able capture boundary, unlike AUC accuracy widely used in banking sector.