Treewise importance

[mnwright]

Implements tree-wise importance, i.e. returns a matrix of variable importance (trees x features). Usage:

rf <- ranger(Species ~ ., data = iris, importance = "permutation", treewise.importance = TRUE)
rf$variable.importance

Also works for other importance modes, e.g.:

rf <- ranger(Species ~ ., data = iris, importance = "impurity_corrected", treewise.importance = TRUE)
rf$variable.importance