apply_rfe: Apply recursive feature elimination

In konradsemsch/aider: Perform everyday analytical and data science tasks more efficiently

Description

This function performes Recursive Feature Elimination (RFE) based on the implementation in the caret package. It is currently only implemented for classification problems. The best subset of variables is selected through maximizing the F1 score. The RFE process is performed by applying bootstrap sampling.

Usage

apply_rfe(df, target, subsets = c(4, 8, 16, 24, 32), number = 25,
  ntree = 500, downsample = "yes")

Arguments

df	A a data frame
target	Target variable
subsets	Provide a vector of the number of variables to fit models to. Defaults to c(4, 8, 16, 24, 32)
number	Number of repeats of the RFE process. Defaults to 25
ntree	Number of random forest trees to fit. Defaults to 500
downsample	Should the majority class be downsampled during resampling? Defaults to "yes"

Examples

data <- recipes::credit_data %>%
  first_to_lower() %>%
  apply_recipe(status) %>%
  prep(retain = TRUE) %>%
  juice()

rfe <- apply_rfe(data, status, number = 10, subsets = c(5, 10, 15))

konradsemsch/aider documentation built on May 22, 2019, 2:40 p.m.