Interpreting Predictions from Complex Models
With large modern datasets the best accuracy is often achieved by complex models even experts struggle to interpret. Here, we present a unified framework for interpreting predictions, namely SHAP (SHapley Additive exPlanations)