How to get results from knn by using recipe, parsnip, tidymodels in r?

Question

I am new in R & getting a bit confused with steps need to be followed in classification task by using tidymodels.

kaggle dataset from: https://www.kaggle.com/c/home-credit-default-risk to perform a Classification on TARGET variable of dataset.

This dataset has both missing & negative values and numeric + categorical data.

ISSUE: Getting error on collect_metrics() after fitting model in step 4

Steps I followed:

library(tidyverse)
library(caret)
library(tidymodels)

After some EDA I have followed below steps:

1. Data Partition - Train / test

set.seed(1234)
data_split <- initial_split(dt1, strata = TARGET)

dt1_train <- training(data_split)
dt1_test <- testing(data_split)

dim(dt1_train)
dim(dt1_test)

########## output ############
[1] 230634    122
[1] 76877   122

2. Recipes

rec2 <- recipe(TARGET ~ ., data = dt1_train) %>% 
    step_rm(contains("SK_ID_CURR")) %>% # removing id var
    step_medianimpute(all_numeric()) %>% 
    step_modeimpute(all_nominal()) %>% 
    step_dummy(all_nominal(), - all_outcomes()) %>% 
    step_range(all_numeric()) %>% # to convert negative numbers into range 0 to 1 
    step_BoxCox(all_numeric()) %>% # as boxcox transformation works on positive numbers only
    step_normalize(all_numeric()) %>% 
    step_zv(all_numeric()) %>% 
    step_nzv(all_numeric()) %>%
    step_corr(all_numeric())

3. RFE

Sampling - to reduce data size for faster rfe results

dt1_train_baked_sample <- bake(prepd_rec2, new_data = dt1_train) %>% sample_frac(0.05)

dim(dt1_train_baked_sample)

######## output ##########
[1] 11532    72

control <- rfeControl(functions = rfFuncs, method = "cv", verbose = FALSE)

system.time(
  RFE_res <- rfe(x = subset(dt1_train_baked_sample, select = -TARGET),
                 y = dt1_train_baked_sample$TARGET, 
                 sizes = c(7, 15, 20),
                 rfeControl = control
                 )
) 

RFE_res$optVariables[1:15] 

######## output ##########
[1] "EXT_SOURCE_2"                                      "EXT_SOURCE_1"                                     
 [3] "DAYS_BIRTH"                                        "AMT_INCOME_TOTAL"                                 
 [5] "CODE_GENDER_M"                                     "DAYS_ID_PUBLISH"                                  
 [7] "AMT_CREDIT"                                        "REG_CITY_NOT_WORK_CITY"                           
 [9] "CNT_FAM_MEMBERS"                                   "AMT_ANNUITY"                                      
[11] "NAME_EDUCATION_TYPE_Higher.education"              "REGION_POPULATION_RELATIVE"                       
[13] "NAME_EDUCATION_TYPE_Secondary...secondary.special" "REG_CITY_NOT_LIVE_CITY"                           
[15] "DAYS_REGISTRATION" 

4. Model Building

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

knn_Spec

knn_fit <- knn_Spec %>%
  
  fit(TARGET ~ EXT_SOURCE_2+EXT_SOURCE_1+DAYS_BIRTH+AMT_INCOME_TOTAL+CODE_GENDER_M+DAYS_ID_PUBLISH+AMT_CREDIT+   REG_CITY_NOT_WORK_CITY+CNT_FAM_MEMBERS+AMT_ANNUITY+NAME_EDUCATION_TYPE_Higher.education+REGION_POPULATION_RELATIVE+NAME_EDUCATION_TYPE_Secondary...secondary.special+REG_CITY_NOT_LIVE_CITY+DAYS_REGISTRATION,
      
      data = dt1_train_baked)

knn_fit

knn_fit %>% collect_metrics()

Error: No `collect_metric()` exists for this type of object

I am not sure how to get results like accuracy, spec, sens & predictions/prob from this.

Also tried below code but that gives an error:

knn_workflow <- workflow() %>% 
  
  add_recipe(rec2) %>% 
  add_model(knn_fit)

Error: `spec` must be a `model_spec`

knn_workflow <- workflow() %>% 
  
  add_recipe(rec2) %>% 
  add_model(knn_Spec)

knn_workflow %>% collect_metrics()

Error: No `collect_metric()` exists for this type of object.