ml_wflowideas.R

library(AmesHousing)
library(tidymodels)

ames <- make_ames()

ml_wflow <-
  ames %>%
  workflow() %>%
  initial_split(prop = .75) %>%
  recipe(Sale_Price ~ Longitude + Latitude + Neighborhood, data = ames) %>%
  step_log(Sale_Price, base = 10) %>%
  step_other(Neighborhood, threshold = 0.05) %>%
  step_dummy(recipes::all_nominal()) %>%
  step_scale(Sale_Price) %>%
  step_scale(Latitude)

mod1 <-
  linear_reg(penalty = tune(), mixture = tune()) %>%
  set_engine("glmnet") %>%
  grid_regular(levels = 10)

mod2 <-
  rand_forest(mtry = tune(), trees = tune()) %>%
  set_engine("glmnet") %>%
  grid_regular(levels = 10)

## The combination of empty parameter specification with any grid_*
## should throw an error.

ml_wflow %>%
  add_model(mod1) %>%
  vfold_cv() %>%
  fit()

# Add new recipes
final_res <-
  ml_wflow %>%
  update_rcp(
    .recipe %>% 
    step_scale(Sale_Price) %>%
    step_center(Sale_Price)
  ) %>%
  add_model(mod2) %>%
  vfold_cv() %>% 
  fit()

## Add new variable
ml_wflow %>%
  update_frm(
    ~ . + x2
  ) %>% 
  update_rcp(
    .rcp %>% step_dummy(x2)
  ) %>%
  add_model(mod2) %>%
  vfold_cv() %>% 
  fit()

## Replace y
ml_wflow %>%
  update_frm(
    Latitude ~ .
  ) %>% 
  update_rcp(
    .rcp %>% step_log(Latitude)
  ) %>%
  add_model(mod2) %>%
  vfold_cv() %>% 
  fit()

## Update whole formula
ml_wflow %>%
  update_frm(
    Latitude ~ Neighborhood + whatever
  ) %>% 
  update_rcp(
    .rcp %>%
    step_log(Latitude) %>%
    step_dummy(Neighborhood) %>%
    step_BoxCox(Whatever),
    # Remove recipe and accept new one
    new = TRUE
  ) %>%
  add_model(mod2) %>%
  vfold_cv() %>% 
  fit()



final_res %>%
  update_rcp(
    -step_scale,
    
  )





# ml_wflow shouldn't run anything -- it's just a specification
# of all the different steps. `fit` should run everything
ml_wflow <- fit(ml_wflow)

# Plot results of tuning parameters
ml_wflow %>%
  autoplot()

# Automatically extract best parameters and fit to the training data
final_model <-
  ml_wflow %>%
  fit_best_model(metrics = metric_set(rmse))

# Predict on the test data using the last model
# Everything is bundled into a workflow object
# and everything can be extracted with separate
# functions with the same verb
final_model %>%
  holdout_error()