Syntax

Here’s a look at the function’s syntax:

fast_regression(
  .data,
  .rec_obj,
  .parsnip_fns = "all",
  .parsnip_eng = "all",
  .split_type = "initial_split",
  .split_args = NULL,
  .drop_na = TRUE
)

Arguments

• .data: The data frame to be used in the regression problem.
• .rec_obj: A recipe object from the recipes package that defines the pre-processing steps.
• .parsnip_fns: Specifies which parsnip functions to use. The default "all" will create all possible regression model specifications.
• .parsnip_eng: Specifies which parsnip engines to use. The default "all" will create all possible regression model specifications.
• .split_type: Defines the type of data split, with "initial_split" as the default. Other split types supported by rsample can also be used.
• .split_args: Additional arguments for the split type. When set to NULL, default parameters for the chosen split type are used.
• .drop_na: Determines whether to drop NA values from the data. Default is TRUE.

Output

The function returns a tibble with details about the generated models:

fast_reg_tbl

# A tibble: 2 × 8
  .model_id .parsnip_engine .parsnip_mode .parsnip_fns model_spec wflw      
      <int> <chr>           <chr>         <chr>        <list>     <list>    
1         1 lm              regression    linear_reg   <spec[+]>  <workflow>
2         2 glm             regression    linear_reg   <spec[+]>  <workflow>
# ℹ 2 more variables: fitted_wflw <list>, pred_wflw <list>

Explanation

• .model_id: A unique identifier for each model.
• .parsnip_engine: The engine used by parsnip (e.g., lm, glm).
• .parsnip_mode: The mode of the model, typically regression.
• .parsnip_fns: The parsnip function used (e.g., linear_reg).
• model_spec: The model specification.
• wflw: The workflow object.
• fitted_wflw: The fitted workflow object.
• pred_wflw: The predictions from the fitted workflow.

Benefits of fast_regression()

• Efficiency: Quickly set up and evaluate multiple regression models with different specifications.
• Flexibility: Supports a wide range of engines and functions, allowing for comprehensive model exploration.
• Integration: Seamlessly integrates with the recipes, parsnip, and rsample packages, making it a versatile tool in your modeling arsenal.

Let’s Inspect

Now that we have our models, let’s take a closer look at the results. We can use the extract_regression_residuals() function to extract the residuals from the fitted models along with the original and predicted data:

extract_regression_residuals(fast_reg_tbl)

[[1]]
# A tibble: 32 × 4
   .model_type     .actual .predicted  .resid
   <chr>             <dbl>      <dbl>   <dbl>
 1 lm - linear_reg    21.4       21.1  0.276 
 2 lm - linear_reg    21         23.9 -2.91  
 3 lm - linear_reg    19.7       18.7  1.01  
 4 lm - linear_reg    18.1       19.7 -1.55  
 5 lm - linear_reg    14.7       12.0  2.73  
 6 lm - linear_reg    24.4       23.7  0.694 
 7 lm - linear_reg    22.8       24.3 -1.48  
 8 lm - linear_reg    13.3       12.5  0.820 
 9 lm - linear_reg    10.4       13.2 -2.85  
10 lm - linear_reg    19.2       19.3 -0.0583
# ℹ 22 more rows

[[2]]
# A tibble: 32 × 4
   .model_type      .actual .predicted  .resid
   <chr>              <dbl>      <dbl>   <dbl>
 1 glm - linear_reg    21.4       21.1  0.276 
 2 glm - linear_reg    21         23.9 -2.91  
 3 glm - linear_reg    19.7       18.7  1.01  
 4 glm - linear_reg    18.1       19.7 -1.55  
 5 glm - linear_reg    14.7       12.0  2.73  
 6 glm - linear_reg    24.4       23.7  0.694 
 7 glm - linear_reg    22.8       24.3 -1.48  
 8 glm - linear_reg    13.3       12.5  0.820 
 9 glm - linear_reg    10.4       13.2 -2.85  
10 glm - linear_reg    19.2       19.3 -0.0583
# ℹ 22 more rows

Now of course we must now visualize! Let’s plot the residuals to see how well our models are performing:

extract_regression_residuals(fast_reg_tbl) |> 
  plot_regression_residuals()

[[1]]

[[2]]

Now the residuals are plotted, we can see how well our models are performing. This is a great way to visually inspect the quality of our models and identify any potential issues.

Let’s look at the predicted vs actual only now:

extract_wflw_pred(fast_reg_tbl,1:nrow(fast_reg_tbl)) |>
  plot_regression_predictions()

[[1]]

[[2]]