Clean_credit <- readRDS("Credit_clean.RDS")

dane_model <- Clean_credit %>%
  select(MIS_Status, State, GrAppv, NAICS, SBA_Appv, NewExist)


library(tidymodels)
set.seed(12345)

credit_split <- initial_split(dane_model, strata = MIS_Status)

credit_train <- training(credit_split)
credit_test <- testing(credit_split)


cred_rec <- recipe(MIS_Status ~ ., data = credit_train) %>%
  step_corr(all_numeric()) %>%
  step_dummy(all_nominal(), -all_outcomes()) %>%
  step_zv(all_numeric()) %>%
  step_normalize(all_numeric()) %>%
  prep()

## GLM
glm_spec <- logistic_reg() %>%
  set_engine("glm")

glm_fit <-  glm_spec %>%
  fit(MIS_Status ~ ., data = juice(cred_rec))

glm_fit

## KNN
#install.packages("kknn")
library(kknn)

knn_spec <- nearest_neighbor() %>%
  set_engine("kknn") %>%
  set_mode("classification")

knn_fit <-  knn_spec %>%
  fit(MIS_Status ~ ., data = juice(cred_rec))

knn_fit

## DECISION TREEES

tree_spec <- decision_tree() %>%
  set_engine("rpart") %>%
  set_mode("classification")

tree_fit <-  tree_spec %>%
  fit(MIS_Status ~ ., data = juice(cred_rec))

tree_fit