Fit a workflow object — fit-workflow • workflows

Fitting a workflow currently involves three main steps:

Preprocessing the data using a formula preprocessor, or by calling recipes::prep() on a recipe.
Fitting the underlying parsnip model using parsnip::fit.model_spec().
Postprocessing predictions from the model using tailor::tailor().

Usage

# S3 method for class 'workflow'
fit(object, data, ..., calibration = NULL, control = control_workflow())

Arguments

object: A workflow
data: A data frame of predictors and outcomes to use when fitting the preprocessor and model.
...: Not used
calibration: A data frame of predictors and outcomes to use when fitting the postprocessor. See the "Data Usage" section of add_tailor() for more information.
control: A control_workflow() object

Value

The workflow object, updated with a fit parsnip model in the object$fit$fit slot.

Indicator Variable Details

Some modeling functions in R create indicator/dummy variables from categorical data when you use a model formula, and some do not. When you specify and fit a model with a workflow(), parsnip and workflows match and reproduce the underlying behavior of the user-specified model’s computational engine.

Formula Preprocessor

In the modeldata::Sacramento data set of real estate prices, the type variable has three levels: "Residential", "Condo", and "Multi-Family". This base workflow() contains a formula added via add_formula() to predict property price from property type, square footage, number of beds, and number of baths:

set.seed(123)

library(parsnip)
library(recipes)
library(workflows)
library(modeldata)

data("Sacramento")

base_wf <- workflow() %>%
  add_formula(price ~ type + sqft + beds + baths)

This first model does create dummy/indicator variables:

lm_spec <- linear_reg() %>%
  set_engine("lm")

base_wf %>%
  add_model(lm_spec) %>%
  fit(Sacramento)

## == Workflow [trained] ================================================
## Preprocessor: Formula
## Model: linear_reg()
##
## -- Preprocessor ------------------------------------------------------
## price ~ type + sqft + beds + baths
##
## -- Model -------------------------------------------------------------
##
## Call:
## stats::lm(formula = ..y ~ ., data = data)
##
## Coefficients:
##      (Intercept)  typeMulti_Family   typeResidential
##          32919.4          -21995.8           33688.6
##             sqft              beds             baths
##            156.2          -29788.0            8730.0

There are five independent variables in the fitted model for this OLS linear regression. With this model type and engine, the factor predictor type of the real estate properties was converted to two binary predictors, typeMulti_Family and typeResidential. (The third type, for condos, does not need its own column because it is the baseline level).

This second model does not create dummy/indicator variables:

rf_spec <- rand_forest() %>%
  set_mode("regression") %>%
  set_engine("ranger")

base_wf %>%
  add_model(rf_spec) %>%
  fit(Sacramento)

## == Workflow [trained] ================================================
## Preprocessor: Formula
## Model: rand_forest()
##
## -- Preprocessor ------------------------------------------------------
## price ~ type + sqft + beds + baths
##
## -- Model -------------------------------------------------------------
## Ranger result
##
## Call:
##  ranger::ranger(x = maybe_data_frame(x), y = y, num.threads = 1,      verbose = FALSE, seed = sample.int(10^5, 1))
##
## Type:                             Regression
## Number of trees:                  500
## Sample size:                      932
## Number of independent variables:  4
## Mtry:                             2
## Target node size:                 5
## Variable importance mode:         none
## Splitrule:                        variance
## OOB prediction error (MSE):       7058847504
## R squared (OOB):                  0.5894647

Note that there are four independent variables in the fitted model for this ranger random forest. With this model type and engine, indicator variables were not created for the type of real estate property being sold. Tree-based models such as random forest models can handle factor predictors directly, and don’t need any conversion to numeric binary variables.

Recipe Preprocessor

When you specify a model with a workflow() and a recipe preprocessor via add_recipe(), the recipe controls whether dummy variables are created or not; the recipe overrides any underlying behavior from the model’s computational engine.

Examples

library(parsnip)
library(recipes)
library(magrittr)

model <- linear_reg() %>%
  set_engine("lm")

base_wf <- workflow() %>%
  add_model(model)

formula_wf <- base_wf %>%
  add_formula(mpg ~ cyl + log(disp))

fit(formula_wf, mtcars)
#> ══ Workflow [trained] ════════════════════════════════════════════════════
#> Preprocessor: Formula
#> Model: linear_reg()
#> 
#> ── Preprocessor ──────────────────────────────────────────────────────────
#> mpg ~ cyl + log(disp)
#> 
#> ── Model ─────────────────────────────────────────────────────────────────
#> 
#> Call:
#> stats::lm(formula = ..y ~ ., data = data)
#> 
#> Coefficients:
#> (Intercept)          cyl  `log(disp)`  
#>     67.6674      -0.1755      -8.7971  
#> 

recipe <- recipe(mpg ~ cyl + disp, mtcars) %>%
  step_log(disp)

recipe_wf <- base_wf %>%
  add_recipe(recipe)

fit(recipe_wf, mtcars)
#> ══ Workflow [trained] ════════════════════════════════════════════════════
#> Preprocessor: Recipe
#> Model: linear_reg()
#> 
#> ── Preprocessor ──────────────────────────────────────────────────────────
#> 1 Recipe Step
#> 
#> • step_log()
#> 
#> ── Model ─────────────────────────────────────────────────────────────────
#> 
#> Call:
#> stats::lm(formula = ..y ~ ., data = data)
#> 
#> Coefficients:
#> (Intercept)          cyl         disp  
#>     67.6674      -0.1755      -8.7971  
#>