Counterfactual Analysis is a powerful tool in Explainable Machine Learning. Given a classifier and a record, one seeks the smallest perturbation necessary to have the perturbed record, called the counterfactual explanation, classified in the desired class.
Feature uncertainty in data reflects the inherent variability and noise present in real-world scenarios, and therefore, there is a need to design Counterfactual Analysis methods that take into account uncertainty and provide robust counterfactual explanations.
In this paper, we address the problem of finding counterfactual solutions for tabular data under uncertainty, where uncertainty is modeled assuming each record has a (convex) set around.
The model is expressed as an optimization problem, which is solved with a Variable Neighborhood Search heuristic.