Shapley Values



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 (see papers for details and citations).

It is powerful feature importance tool that is gathering a lot of attention in both the machine learning and quantitative finance community.

There is a really good, open source (MIT License) implementation of the technique, available on Github.


Presentation Slides

../_images/shap.png