CODEWITHSUNDEEP
rpcatidymodels
Manu
Manu

Reputation: 1120

How can I get PCA cumulative proportion with Tidymodels?

I can apply PCA on the classic Iris dataset to obtain the cumulative proportion per dimension:

library(tidyverse)
x <- iris[,1:4] %>% as.matrix()
pca <- prcomp(x)
summary(pca)

But I don't know how can I do that with tidymodels. My code so far is:

library(tidymodels)
iris_vars <- iris %>% select(-Species)
iris_rec <- recipe(~., iris_vars) %>%
  step_pca(all_predictors())
iris_prep <- prep(iris_rec)
iris_tidy <- tidy(iris_prep,1)
iris_tidy
summary(iris_tidy)

I would like to obtain this with tidymodels:

Importance of components:
                          PC1     PC2    PC3     PC4
Standard deviation     2.0563 0.49262 0.2797 0.15439
Proportion of Variance 0.9246 0.05307 0.0171 0.00521
Cumulative Proportion  0.9246 0.97769 0.9948 1.00000

Any help will be greatly appreciated.

Upvotes: 3

Views: 577

Answers (1)

hplieninger
hplieninger

Reputation: 3494

You can get the same results, if you use the same model. prcomp() defaults to center = TRUE, whereas step_pca() defaults to center = FALSE. In the following, I use centering and scaling for both (since this is often recommended).

library("tidymodels")

x <- iris[,1:4] %>% as.matrix()
pca <- prcomp(x, scale. = TRUE)
summary(pca)
#> Importance of components:
#>                           PC1    PC2     PC3     PC4
#> Standard deviation     1.7084 0.9560 0.38309 0.14393
#> Proportion of Variance 0.7296 0.2285 0.03669 0.00518
#> Cumulative Proportion  0.7296 0.9581 0.99482 1.00000

iris_rec <- recipe(Species ~ ., iris) %>%
    step_normalize(all_predictors()) %>% 
    step_pca(all_predictors())
iris_prep <- prep(iris_rec)

summary(iris_prep$steps[[2]]$res)
#> Importance of components:
#>                           PC1    PC2     PC3     PC4
#> Standard deviation     1.7084 0.9560 0.38309 0.14393
#> Proportion of Variance 0.7296 0.2285 0.03669 0.00518
#> Cumulative Proportion  0.7296 0.9581 0.99482 1.00000

Created on 2020-05-29 by the reprex package (v0.3.0)

Upvotes: 4

Related Questions