Selecting based on correlation with other features

Dimensionality Reduction in R

Matt Pickard

Owner, Pickard Predictives, LLC

Review correlation plot creation

healthcare_df %>% 
  select(where(is.numeric)) %>%

  correlate() %>%

  shave() %>%

  rplot(print_cor = TRUE) +

  theme(axis.text.x = element_text(angle = 90, hjust = 1))

Correlation plot

correlation plot of the healthcare company attrition

Correlation strength

correlation plot of the healthcare company attrition with table of correlation strength ranges

A correlation filter?

correlation plot of the healthcare company attrition

A correlation filter?

venn diagram of percent salary hike and performance rating showing large overlap of mutual information

A correlation filter?

Venn diagram of percent salary hike with performance rating removed

A correlation filter?

Both parts of the Venn diagram removed to show we removed valuable information

A correlation filter recipe

# create and prep the recipe
corr_recipe <-  
  recipe(Attrition ~ ., data = healthcare_df) %>%

  step_corr(all_numeric_predictors(), threshold = 0.7) %>%

  prep()


# Apply the recipe to the data
filtered_healthcare_df <- 
  corr_recipe %>% 
  bake(new_data = NULL)


# Identify the features that were removed
tidy(corr_recipe, number = 1)

Let's practice!

Dimensionality Reduction in R