In jbkunst/risk3r: Tools for credit risk modelling

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(risk3r)
library(scorecard)
library(dplyr)
library(ggplot2)

theme_set(
  theme_minimal(base_size = 21) + 
    theme(axis.text.x = element_text(angle = 45, hjust = 1))
  )

if(require(showtext)){

  sysfonts::font_add_google("IBM Plex Sans", "plex")
  showtext::showtext_auto()

}

data("germancredit")

`woebin_summary`

bins <- woebin(germancredit, y = "creditability")

binssummary <- woebin_summary(bins)

binssummary

binssummary %>% 
  select(variable, ks, iv, iv_lbl,
         hhi, hhi_lbl, distribution)

`woebin_ply_min`

The woebin_ply_min function don't need a data frame argument.

variable <- head(germancredit$credit.amount, 6)

bin <- bins$credit.amount

bin

woebin_ply_min(variable, bin)

woebin_ply_min(variable, bin, value = "posprob")

`woebin_cor_iv`

datcor <- woebin_cor_iv(germancredit, bins)

datcor

cor_limit <- 0.15

datcor %>%
  filter(var1 != var2) %>%
  mutate(
    cor_conflict = ifelse(abs(r) > cor_limit, TRUE, FALSE),
    variable_to_remove = ifelse(
      cor_conflict,
      ifelse(var1 > var2, var2, var1),
      NA
    )
  )