Hi Dylan Smith,

For Random Forest regression models, feature importance values are determined from variance reduction (Spark & Scikit-learn calculate this with mean square error).

Gini impurity, or alternatively Entropy, is used for Random Forest classification models.

Best,

Stacey