library(tidymodels)
library(tune)
library(AmesHousing)
library(workflows)
# ------------------------------------------------------------------------------
ames <- make_ames()
# Make sure that you get the same random numbers
set.seed(4595)
data_split <- initial_split(ames, strata = "Sale_Price")
ames_train <- training(data_split)
set.seed(2453)
cv_splits <- vfold_cv(ames_train, v = 10, strata = "Sale_Price")
# ------------------------------------------------------------------------------
ames_rec <-
recipe(Sale_Price ~ ., data = ames_train) %>%
step_log(Sale_Price, base = 10) %>%
step_YeoJohnson(Lot_Area, Gr_Liv_Area)
cart_model <-
decision_tree(
mode = "regression", cost_complexity = tune(), min_n = tune()) %>%
set_engine("rpart")
ames_wflow <-
workflow() %>%
add_recipe(ames_rec) %>%
add_model(cart_model)
extr <- function(x) {
# tibble(num_nods = sum(x$model$frame$var == "<leaf>"))
x$model
}
num_leaves <- function(x) {
tibble(num_leaves = sum(x$model$frame$var == "<leaf>"))
}
prm <-
parameters(ames_wflow) %>%
update(min_n = min_n(c(3, 10)))
set.seed(4567367)
initial_grid <-
tune_grid(
ames_wflow,
resamples = cv_splits,
param_info = prm,
control = control_grid(verbose = TRUE, extract = num_leaves)
)
# ------------------------------------------------------------------------------
foo <- function(i) {
expo_decay(i, start_val = .02, 0, slope = 1/10)
}
set.seed(463)
test <-
tune_bayes(
ames_wflow,
resamples = cv_splits,
initial = initial_grid,
metrics = metric_set(rmse),
objective = exp_improve(foo),
param_info = prm,
iter = 20,
control = control_bayes(verbose = TRUE, uncertain = 10, extract = num_leaves)
)
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.