oblique random forests for classification and regression

[bcjaeger]

Hello!

aorsf has recently been updated to allow for oblique classification and regression forests. May I submit a PR that would add a classification and regression mode for the aorsf engine?

There are a few datasets where the oblique random forest is really helpful (e.g., modeldata::meats)

suppressPackageStartupMessages({
  library(modeldata)
  library(rsample)
  library(recipes)
  library(workflows)
  library(workflowsets)
  library(yardstick)
})
#> Warning: package 'modeldata' was built under R version 4.3.3
#> Warning: package 'yardstick' was built under R version 4.3.3

# load my branch
devtools::load_all(path = "D:/parsnip/")
#> ℹ Loading parsnip

meat_rec <- 
  recipe(protein ~ ., data = meats) %>%
  step_select(-water, -fat)

meat_folds <- vfold_cv(meats)

meat_models <- list(oblique = rand_forest(mode = 'regression', 
                                          engine = 'aorsf'),
                    axis = rand_forest(mode = 'regression',
                                       engine = 'ranger'),
                    xgb = boost_tree(mode = 'regression', 
                                     engine = 'xgboost',
                                     trees = 500))

workflows <- workflow_set(list(meat_rec), meat_models, cross = TRUE)

res <- workflows %>% 
  workflow_map("fit_resamples", 
               verbose = TRUE,
               resamples = meat_folds,
               metrics = metric_set(rsq))
#> i 1 of 3 resampling: recipe_oblique
#> ✔ 1 of 3 resampling: recipe_oblique (3.8s)
#> i 2 of 3 resampling: recipe_axis
#> ✔ 2 of 3 resampling: recipe_axis (2.1s)
#> i 3 of 3 resampling: recipe_xgb
#> ✔ 3 of 3 resampling: recipe_xgb (6.2s)

collect_metrics(res)
#> # A tibble: 3 × 9
#>   wflow_id       .config    preproc model .metric .estimator  mean     n std_err
#>   <chr>          <chr>      <chr>   <chr> <chr>   <chr>      <dbl> <int>   <dbl>
#> 1 recipe_oblique Preproces… recipe  rand… rsq     standard   0.944    10 0.00858
#> 2 recipe_axis    Preproces… recipe  rand… rsq     standard   0.529    10 0.0582 
#> 3 recipe_xgb     Preproces… recipe  boos… rsq     standard   0.524    10 0.0574

^{Created on 2024-05-03 with reprex v2.1.0}

[simonpcouch]

Duplicate of https://github.com/tidymodels/bonsai/issues/73. Closing so as not to track duplicate issue, but we're certainly interested in making this happen!

Looks like you've got an implementation put together locally? I'd be more than happy to work with you to get this merged into bonsai if you're game to start a PR over there. :)

[bcjaeger]

Thank you! I didn't know about bonsai, but it looks awesome. =] I will open a PR there soon.

[github-actions[bot]]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.