CODEWITHSUNDEEP
rrandom-foresthyperparameterstidymodelsr-parsnip
Lillian Welsh
Lillian Welsh

Reputation: 39

creating a tuning grid for Regression Trees in R

Attempting my first randomForest model in R and am working through tuning hyperparameters. I created a spec first: tune_spec<- decision_tree() %>% set_engine("rpart") %>% set_mode("regression")

And then I tried to create a tuning grid: tree_grid<- grid_regular(parameters(tune_spec), levels=3)

but I got this error: Error in parameters(): ! parameters objects cannot be created from objects of class decision_tree.

Upvotes: 0

Views: 281

Answers (1)

EmilHvitfeldt
EmilHvitfeldt

Reputation: 3185

The parameters() function was deprecated in 0.2.0 of {tune} (2022-03-18). The function to use is extract_parameter_set_dials().

library(tidymodels)

tune_spec <- decision_tree() |>
  set_engine("rpart") |>
  set_mode("regression")

tune_spec |> 
  extract_parameter_set_dials()
#> Collection of 0 parameters for tuning
#> 
#> [1] identifier type       object    
#> <0 rows> (or 0-length row.names)

We are getting back a empty parameters object, because you need to specify which variables you want to use with tune()

tune_spec <- decision_tree(tree_depth = tune(), min_n = tune()) |>
  set_engine("rpart") |>
  set_mode("regression")

tune_spec |> 
  extract_parameter_set_dials()
#> Collection of 2 parameters for tuning
#> 
#>  identifier       type    object
#>  tree_depth tree_depth nparam[+]
#>       min_n      min_n nparam[+]

And once you have done that, then you can pass it to grid_regular() or the other functions

tune_spec |> 
  extract_parameter_set_dials() |>
  grid_regular(levels = 3)
#> # A tibble: 9 × 2
#>   tree_depth min_n
#>        <int> <int>
#> 1          1     2
#> 2          8     2
#> 3         15     2
#> 4          1    21
#> 5          8    21
#> 6         15    21
#> 7          1    40
#> 8          8    40
#> 9         15    40

Upvotes: 0

Related Questions