Explainable machine learning with mlr3 and DALEX
Przemysław Biecek and Szymon Maksymiuk added a new chapter to the mlr3 book on how to analyze machine learning models fitted with mlr3 using the excellent DALEX package.
The contributed chapter covers an analysis of a random regression forest (implemented in the ranger package) on data extracted from the FIFA video game. In more detail, the following methods for explainable machine learning are showcased:
- Dataset level exploration: Feature importance and Partial dependency plots.
- Instance level explanation: Break Down, SHapley Additive exPlanations (SHAP), and Ceteris Paribus plots.
Here is a small preview illustrating the effect of different features on the monetary value of Cristiano Ronaldo:
Read the complete chapter here.