Error in UseMethod("conf_mat") : no applicable method for 'conf_mat' applied to an object of class "list"

Question

I am currently trying to create a TidyModel logistic regression model on a bankruptcy dataset. I've been using Rebecca Barters guide to create the setup. This is my first model like this, so any help is appreciated.

When trying to create the conf. matrix I get the following error: Error in UseMethod("conf_mat") : no applicable method for 'conf_mat' applied to an object of class "list"

1. Can anyone help me convert this from a list? I can't seem to do it to get the conf.matrix running.
2. When running the "model_performance" line I get an rmse and rsq score, but I expected to get accuracy and roc_auc.

Thanks a lot.

This is the head of my initial data frame "debt_data": structure(list(bankrupt = c(1L, 1L, 1L, 1L, 1L, 1L), Interest.bearing.debt.interest.rate = c(0.000725072507250725, 0.000647064706470647, 0.00079007900790079, 0.000449044904490449, 0.000686068606860686, 0.000716071607160716), total_debt_vs_total_net_worth = c(0.0212659243655332, 0.0125023937843679, 0.0212476860084444, 0.00957240171805324, 0.00514960012402083, 0.0142131516792967), debt_ratio = c(0.207576261450555, 0.171176346101521, 0.207515796474892, 0.151464764035432, 0.106509054630105, 0.180427487377936)), row.names = c(NA, 6L), class = "data.frame")

This is the code:

debt_data <- data %>% 
  select(
    Bankrupt.,
    Interest.bearing.debt.interest.rate,
    Total.debt.Total.net.worth,
    Debt.ratio..,
  ) %>% 
  rename(
    bankrupt = Bankrupt.,
    total_debt_vs_total_net_worth = Total.debt.Total.net.worth,
    debt_ratio = Debt.ratio..
  )

set.seed(234589)
debt_split <- initial_split(debt_data, 
                             prop = 3/4)
debt_split 

debt_train <- training(debt_split)
debt_test <- testing(debt_split)

model1_cv <- vfold_cv(debt_test)

debt_recipe <- recipe(bankrupt ~
                      Interest.bearing.debt.interest.rate +
                      total_debt_vs_total_net_worth +
                      debt_ratio, 
                      data = debt_data) %>% 
                      step_normalize(all_numeric_predictors()) %>%
                      step_impute_knn(all_predictors())
debt_recipe

model_workflow <- workflow() %>%
  add_recipe(debt_recipe) %>%
  add_model(linear_reg())


model_fit <- model_workflow %>% 
  last_fit(debt_split)
model_fit

model_performance <- model_fit %>% collect_metrics()
model_performance

model_predictions <- model_fit %>% collect_predictions()
model_predictions

model_predictions <- model_fit %>% pull(.predictions)
model_predictions

# Conf. matrix
model_predictions %>% 
  conf_mat(truth = bankrupt, estimate = .pred_class)

I've tried making changes to the model and Googled around, but I can't seem to get any further.

Answer 1

Ulrik, your code with some changes:

1. Used mtcars data to give a reproducible example.

2. Conversion of bankrupt to factor, so you can classify.

3. Full model specification, including set to classify.

4. Removal of conversion of model_predictions to list, as conf_mat takes a tibble (this removes the error).

    library(tidyverse)
    library(tidymodels)
    debt_data <- tibble(bankrupt = if_else(mtcars$cyl == 8, 1, 0) |> as_factor(),
                        Interest.bearing.dept.interest.rate = mtcars$disp,
                        total_debt_vs_total_net_worth = mtcars$disp,
                        debt_ratio = mtcars$hp)
   
    debt_split <- initial_split(debt_data, 
                                prop = 3/4)
    debt_split 
   
    debt_train <- training(debt_split)
    debt_test <- testing(debt_split)
   
    model1_cv <- vfold_cv(debt_train) # Changed from test
   
    debt_recipe <- recipe(bankrupt ~
                            Interest.bearing.dept.interest.rate +
                            total_debt_vs_total_net_worth +
                            debt_ratio, 
                          data = debt_data) %>% 
      step_normalize(all_numeric_predictors()) %>%
      step_impute_knn(all_predictors())
   
    debt_recipe
   
    logistic_reg_glm_spec <-
      logistic_reg(penalty = 1) |> # randomly chosen penalty
      set_engine('glmnet') |>
      set_mode("classification")
   
    model_workflow <- workflow() %>%
      add_recipe(debt_recipe) %>%
      add_model(logistic_reg_glm_spec)
   
    model_workflow
   
    model_fit <- fit(model_workflow, data = debt_train)
   
    model_fit <- model_workflow %>% 
      last_fit(debt_split)
    model_fit
   
    model_performance <- model_fit %>% collect_metrics()
    model_performance
    # Now you get accuracy and roc_auc
   
    model_predictions <- model_fit %>% collect_predictions()
    model_predictions
   
    model_predictions |>
    conf_mat(truth = bankrupt,
           estimate = .pred_class)
   

You were very close, only three small changes! Hope this helps :-)