0

我想在配方中计算出第 X 个 PCA 组件后选择它们。然后我想在工作流程中添加这个配方。

请参阅下面的示例数据。

library(tidymodels)
x1 <- c(1, 6, 4, 2, 3, 4, 5, 7, 8, 2)
x2 <- c(1, 3, 4, 2, 3, 4, 5, 7, 8, 2)
x3 <- c(1, 3, 4, 2, 3, 4, 5, 7, 8, 2)
x4 <- c(1, 3, 4, 2, 3, 4, 5, 7, 8, 2)
id <- c(1:10)
y <- c(1, 4, 2, 5, 6, 2, 3, 6, 2, 4)
df1_train <- tibble(x1, x2,  x3,  x4, id, y)

step_PCA_PREPROCESSING = 4
selectXfirstPCA = 3

# My recipe
df1_train_recipe <- df1_train %>%
  recipes::recipe(y ~ .) %>%
  recipes::update_role(id, new_role = "id variable") %>%
  recipes::step_pca(., recipes::all_predictors(), num_comp = step_PCA_PREPROCESSING) %>% 
  recipes::update_role(tidyselect::num_range("PC", 1:selectXfirstPCA), new_role = "predictor")


# To then continue like below: 

# Model specifications
model_spec <- parsnip::linear_reg() %>% 
  parsnip::set_engine("glmnet") 

# Create workflow (to know variable roles from recipes)
df1_workflow <- workflows::workflow() %>%
  workflows::add_recipe(df1_train_recipe) %>%
  workflows::add_model(model_spec) 

# Fit model
mod <-  parsnip::fit(df1_workflow, data = df1_train)
4

1 回答 1

1

这只是num_comp. step_pca()根据文档，num_comp是“作为新预测变量保留的 PCA 组件的数量。如果 num_comp 大于列数或可能组件的数量，将使用较小的值。”
所以你的代码可以简单地
recipes::step_pca(., recipes::all_predictors(), num_comp = selectXfirstPCA)删除最后update_role()一行。

于 2020-12-14T17:54:27.013 回答