fourth tutorial from tidymodels

04_tune_model_parameters.R

#################################################################
##                          tidymodels                         ##
##                  4  Tune model parameters                  ##
##        url: https://www.tidymodels.org/start/tuning/        ##
#################################################################

# 1.0 INTRODUCTION ----
        ## 1.1 hyperparameters ----
                ### 1.1.1 mtry()----
                ### 1.1.2 learn_rate()----
library(tidymodels)  # for the tune package, along with the rest of tidymodels
# Helper packages
library(modeldata)   # for the cells data
library(vip)         # for variable importance plots
# 2.0 THE CELL IMAGE DATA, REVISITED ----
#labeled by experts as well-segmented (WS) or poorly segmented (PS).
data(cells, package = "modeldata")
# 3.0 PREDICTING IMAGE SEGMENTATION, BUT BETTER ----
        #Random forest models are a tree-based ensemble method, and typically
        #perform well with default hyperparameters. However, the accuracy of
        #some other tree-based models, such as boosted tree models or decision
        #tree models, can be sensitive to the values of hyperparameters. In
        #this article, we will train a decision tree model.
        ## 3.1 cost_complexity() ----
                #adds a cost, or penalty, to error rates of more complex trees
        ## 3.2 tree_depth() ----
                #helps by stopping our tree from growing after it reaches
                #a certain depth.
set.seed(123)
cell_split <- initial_split(cells %>% select(-case),
                            strata = class)
cell_train <- training(cell_split)
cell_test  <- testing(cell_split)
# 4.0 TUNING HYPERPARAMETERS ----
        ## 4.1 identifies which hyperparameters we plan to tune ----
tune_spec <-
        decision_tree(
                cost_complexity = tune(),
                tree_depth = tune()
        ) %>%
        set_engine("rpart") %>%
        set_mode("classification")
tune_spec
        ## 4.2 dials::grid_regular() ----
tree_grid <- grid_regular(cost_complexity(),
                          tree_depth(),
                          levels = 5)
tree_grid
## 4.3 Cross validation folds ----
set.seed(234)
cell_folds <- vfold_cv(cell_train)
# 5.0 MODEL TUNING WITH A GRID ----
        ## 5.1 workflow -----
        set.seed(345)
        tree_wf <- workflow() %>%
                add_model(tune_spec) %>%
                add_formula(class ~ .)
        ## 5.2 add tuning grid ----
        tree_res <-
                tree_wf %>%
                tune_grid(
                        resamples = cell_folds,
                        grid = tree_grid
                )
        tree_res
        ## 5.3 Plot ----
        tree_res %>%
                collect_metrics() %>%
                mutate(tree_depth = factor(tree_depth)) %>%
                ggplot(aes(cost_complexity, mean, color = tree_depth)) +
                geom_line(size = 1.5, alpha = 0.6) +
                geom_point(size = 2) +
                facet_wrap(~ .metric, scales = "free", nrow = 2) +
                scale_x_log10(labels = scales::label_number()) +
                scale_color_viridis_d(option = "plasma", begin = .9, end = 0)
        ## 5.4 Select Best ----
        best_tree <- tree_res %>%
                select_best("roc_auc")
        best_tree

# 6.0 FINALIZING OUR MODEL ----
        final_wf <-
                tree_wf %>%
                finalize_workflow(best_tree)
        final_wf


# 7.0 EXPLORING RESULTS ----
        ## 7.1 fit final model to training data ----
        final_tree <-
                final_wf %>%
                fit(data = cell_train)

        final_tree
        ## 7.2 variable importance ----
        library(vip)
        final_tree %>%
                pull_workflow_fit() %>%
                vip(geom = "point")
# 8.0 LAST FIT ----
        ## 8.1 test data ----
        final_fit <-
                final_wf %>%
                last_fit(cell_split)
        ## 8.2 collect metrics ----
        final_fit %>%
                collect_metrics()
        ## 8.3 plot roc curve ----
        final_fit %>%
                collect_predictions() %>%
                roc_curve(class, .pred_PS) %>%
                autoplot()
        ## 8.4 Other hyperparameters?? ----
        args(decision_tree)
        ## 8.5 https://www.tidymodels.org/find/parsnip/#models ----