在随机森林 tidymodels r 中设置调整游侠的最大深度

Question

I would like to tune the depth of my random forest to avoid overfitting. I am using tidymodels and this is my model code.

rf_model <- rand_forest(mtry = tune(), 
                        trees = tune(),
                        max.depth = tune()) %>%
  set_mode("classification") %>%
  set_engine("ranger", importance = "impurity")

It gives me an error that:

Error in rand_forest(mtry = tune(), trees = tune(), max.depth = tune()): unused argument (max.depth = tune())

I also tried tree_depth = tune() from dials documentation, and that gives the same error.

But when I look at ranger documentation, it has max.depth as a parameter. wondering how to tune depth with tidymodels tune.

Thank you

Answer 1

There is no max.depth argument, so like in ranger (see What is equivalent of "max depth" in the 'R' package "ranger"? for explanation) the minimum number of nodes can be used instead. This works:

rf_model <- rand_forest(mtry = tune(), trees = tune(), min_n = tune()) %>%
    set_mode("classification") %>% set_engine("ranger", importance = "impurity")

Which produces a valid rf_model :

> rf_model
Random Forest Model Specification (classification)

Main Arguments:
  mtry = tune()
  trees = tune()
  min_n = tune()

Engine-Specific Arguments:
  importance = impurity

Computational engine: ranger