Merge pull request #17 from jcierocki/dummy-vars

Dummy variables for Geography

dataset_prep.R

@@ -4,6 +4,7 @@ library(tidyverse)
 library(tidymodels)
 library(stringr)
 library(scorecard)
+library(recipes)
 
 rm(list = ls())
 
@@ -17,15 +18,25 @@ data1 <- data_raw %>%
          HasCrCard = factor(HasCrCard) %>% `levels<-`(c("No", "Yes"))) %>% 
   dplyr::select(-RowNumber, -CustomerId, -Surname)
 
-data1 %>% filter_vars_by_iv(significance_thres = 0.02) %>% 
+data2 <- recipe(Exited ~ ., data = data1) %>% 
+  step_dummy(Geography) %>% 
+  prep %>% bake(new_data = data1)
+
+changed_cols_idx <- data2 %>% colnames %>% str_split("_") %>% map_lgl(~ .x[1] == "Geography")
+changed_cols <- colnames(data2)[changed_cols_idx]
+data2 <- data2 %>% 
+  mutate_at(changed_cols, ~ as.factor(.x) %>% `levels<-`(c("No", "Yes"))) %>% 
+  rename_at(changed_cols, ~ str_remove(.x, "_"))
+
+data2 %>% filter_vars_by_iv(significance_thres = 0.01) %>%
   initial_split(prop = 0.75) %>%
   saveRDS("data/split_raw.RDS")
 
-data2 <- data1 %>% 
+data3 <- data2 %>% 
   factorize(bin_methods = "tree") %>% 
   as_tibble() %>% 
-  filter_vars_by_iv(significance_thres = 0.02)
+  filter_vars_by_iv(significance_thres = 0.01)
 
-dataset_split <- initial_split(data2, prop = 0.75) %>% saveRDS("data/split.RDS")
+dataset_split <- data3 %>% initial_split(prop = 0.75) %>% saveRDS("data/split.RDS")
 
 rm(list = ls())

funs_preproc.R

@@ -41,7 +41,7 @@ choose_best_binning <- function(binnings_df) {
 }
 
 factorize <- function(df, y_name = "Exited", y_pos = "No", bin_limit = 6, bin_methods = c("tree", "chimerge")) {
-  fct_cols <- colnames(df)[data1 %>% map_lgl(~ !is.factor(.x)) & colnames(df) != y_name]
+  fct_cols <- colnames(df)[(df %>% map_lgl(~ !is.factor(.x))) & colnames(df) != y_name]
   binnings <- bin_methods %>% 
     map(~ df %>% woebin(y = y_name, x = fct_cols, positive = y_pos, bin_num_limit = bin_limit, method = .x)) %>% 
     `names<-`(bin_methods) %>% 
@@ -51,7 +51,7 @@ factorize <- function(df, y_name = "Exited", y_pos = "No", bin_limit = 6, bin_me
 
   df %>% woebin_ply(bins = bins_best, to = "bin") %>% 
     mutate_if(~ !is.factor(.x), as.factor) %>% 
-    rename_all(function(x) map_chr(str_split(x, "_"), ~ .x[1])) %>% 
+    rename_all(function(x) str_split(x, "_") %>% map_chr(~ .x[1])) %>% 
     return
 }