如何在 r 中使用 xgboost 和 decision_tree 绘制树？

Question

我试图在使用 tidymodels 工作流程后在 R 中绘制决策树，但我无法找到使用 and/or 好的模型的好函数。在这样的代码之后，你如何编写一个情节？

xgboost_spec <- 
  boost_tree(trees = tune(), min_n = tune(), tree_depth = tune(), learn_rate = tune(), 
             loss_reduction = tune(), sample_size = tune()) %>% 
  set_mode("classification") %>% 
  set_engine("xgboost") 

xgboost_workflow <- 
  workflow() %>% 
  add_recipe(data_recipe) %>% 
  add_model(xgboost_spec) 

xgboost_tune <-
  tune_grid(xgboost_workflow, resamples = data_folds, grid = 10)

final_xgboost <- xgboost_workflow %>% 
  finalize_workflow(select_best(xgboost_tune, "roc_auc"))

xgboost_results <- final_xgboost %>% 
  fit_resamples(
    resamples = data_folds,
    metrics = metric_set(roc_auc, accuracy, sensitivity, specificity),
    control = control_resamples(save_pred = TRUE)
  )

还是在决策树代码之后？

tree_spec <- decision_tree(
  cost_complexity = tune(),
  tree_depth = tune(),
  min_n = tune()) %>% 
  set_engine("rpart") %>% 
  set_mode("classification")

tree_workflow <- 
  workflow() %>% 
  add_recipe(data_recipe) %>% 
  add_model(tree_spec) 

tree_grid <- grid_regular(cost_complexity(),
                          tree_depth(),
                          min_n(), levels = 4)

tree_tune <- tree_workflow %>% 
  tune_grid(
  resamples = data_folds,
  grid = tree_grid,
  metrics = metric_set(roc_auc, accuracy, sensitivity, specificity)
)
final_tree <- tree_workflow %>% 
  finalize_workflow(select_best(tree_tune, "roc_auc"))

tree_results <- final_tree %>% 
  fit_resamples(
    resamples = data_folds,
    metrics = metric_set(roc_auc, accuracy, sensitivity, specificity),
    control = control_resamples(save_pred = TRUE)
  )

可能吗？或者我应该在last_fit()之后使用模型？

谢谢！

Answer 1

我认为绘制 xgboost 模型没有多大意义，因为它是增强树（很多很多树），但您可以绘制单个决策树。

关键是大多数用于可视化树结果的软件包都要求您 repair the call object。

library(tidymodels)
#> Registered S3 method overwritten by 'tune':
#>   method                   from   
#>   required_pkgs.model_spec parsnip

data(penguins)
penguins <- na.omit(penguins)

cart_spec <-
   decision_tree() %>%
   set_engine("rpart") %>%
   set_mode("classification")

cart_fit <- 
   cart_spec %>%
   fit(sex ~ species + bill_length_mm + body_mass_g, data = penguins)
cart_fit <- repair_call(cart_fit, data = penguins)

library(rattle)
#> Loading required package: bitops
#> Rattle: A free graphical interface for data science with R.
#> Version 5.4.0 Copyright (c) 2006-2020 Togaware Pty Ltd.
#> Type 'rattle()' to shake, rattle, and roll your data.
fancyRpartPlot(cart_fit$fit)

^{由 reprex package (v2.0.0)}

于 2021-08-07 创建

rattle package isn't the only thing out there; ggparty 是另一个不错的选择。

这确实意味着您必须使用防风草模型和预处理器，而不是工作流程。可以看到一个tutorial of how to tune a parsnip plus preprocessor here.

如何在 r 中使用 xgboost 和 decision_tree 绘制树？

How to plot a tree with xgboost and decision_tree in r?

tree

workflow

plot

xgboost

tidymodels