SHAP (SHapley Additive exPlanations) is a game theoretic approach to explain
the output of any machine learning model. It connects optimal credit
allocation with local explanations using the classic Shapley values from game
theory and their related extensions.