calc_model_stats <- function(x) {
glance(extract_fit_parsnip(x)) |>
select(adj.r.squared, AIC, BIC)
}AE 11: Cross validation
Important
Go to the course GitHub organization and locate your ae-11 repo to get started.
Render, commit, and push your responses to GitHub by the end of class. The responses are due in your GitHub repo no later than Thursday, October 26 at 11:59pm.
Model statistics function
You will use this function to calculate
Packages
library(tidyverse)
library(tidymodels)
library(knitr)Load data and relevel factors
tips <- read_csv("data/tip-data.csv")
tips <- tips |>
mutate(Age = factor(Age, levels = c("Yadult", "Middle", "SenCit")),
Meal = factor(Meal, levels = c("Lunch", "Dinner", "Late Night"))
)Split data into training and testing
Split your data into testing and training sets.
set.seed(10232023)
tips_split <- initial_split(tips)
tips_train <- training(tips_split)
tips_test <- testing(tips_split)Specify model
Specify a linear regression model. Call it tips_spec.
tips_spec <- linear_reg() |>
set_engine("lm")
tips_specLinear Regression Model Specification (regression)
Computational engine: lm Model 1
Create recipe
Create a recipe to use Party, Age, and Meal to predict Tip. Call it tips_rec1.
tips_rec1 <- recipe(Tip ~ Party + Age + Meal,
data = tips_train) |>
step_dummy(all_nominal_predictors()) |>
step_zv(all_predictors())
tips_rec1── Recipe ──────────────────────────────────────────────────────────────────────── Inputs Number of variables by roleoutcome: 1
predictor: 3── Operations • Dummy variables from: all_nominal_predictors()• Zero variance filter on: all_predictors()Preview recipe
prep(tips_rec1) |>
bake(tips_train) |>
glimpse()Rows: 126
Columns: 6
$ Party <dbl> 3, 2, 2, 4, 2, 7, 4, 3, 2, 4, 1, 2, 2, 1, 2, 1, 2, 3, …
$ Tip <dbl> 4.00, 4.92, 5.09, 8.84, 3.09, 15.00, 8.00, 4.00, 5.00,…
$ Age_Middle <dbl> 1, 0, 1, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 1, 1, …
$ Age_SenCit <dbl> 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, …
$ Meal_Dinner <dbl> 0, 1, 1, 1, 0, 1, 0, 1, 0, 0, 0, 1, 1, 0, 1, 0, 1, 0, …
$ Meal_Late.Night <dbl> 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 1, …Create workflow
Create the workflow that brings together the model specification and recipe. Call it tips_wflow1.
tips_wflow1 <- workflow() |>
add_model(tips_spec) |>
add_recipe(tips_rec1)
tips_wflow1══ Workflow ════════════════════════════════════════════════════════════════════
Preprocessor: Recipe
Model: linear_reg()
── Preprocessor ────────────────────────────────────────────────────────────────
2 Recipe Steps
• step_dummy()
• step_zv()
── Model ───────────────────────────────────────────────────────────────────────
Linear Regression Model Specification (regression)
Computational engine: lm Cross validation
Create folds
Create 5 folds.
# make 10 folds
set.seed(10232023)
folds <- vfold_cv(tips_train, v = 5)Conduct cross validation
Conduct cross validation on the 5 folds.
# Fit model and performance statistics for each iteration
tips_fit_rs1 <- tips_wflow1 |>
fit_resamples(resamples = folds,
control = control_resamples(extract = calc_model_stats))Take a look at tips_fit_rs1
tips_fit_rs1# Resampling results
# 5-fold cross-validation
# A tibble: 5 × 5
splits id .metrics .notes .extracts
<list> <chr> <list> <list> <list>
1 <split [100/26]> Fold1 <tibble [2 × 4]> <tibble [0 × 3]> <tibble [1 × 2]>
2 <split [101/25]> Fold2 <tibble [2 × 4]> <tibble [0 × 3]> <tibble [1 × 2]>
3 <split [101/25]> Fold3 <tibble [2 × 4]> <tibble [0 × 3]> <tibble [1 × 2]>
4 <split [101/25]> Fold4 <tibble [2 × 4]> <tibble [0 × 3]> <tibble [1 × 2]>
5 <split [101/25]> Fold5 <tibble [2 × 4]> <tibble [0 × 3]> <tibble [1 × 2]>Summarize assessment CV metrics
Summarize assessment metrics from your CV iterations These statistics are calculated using the assessment set.
collect_metrics(tips_fit_rs1, summarize = TRUE)# A tibble: 2 × 6
.metric .estimator mean n std_err .config
<chr> <chr> <dbl> <int> <dbl> <chr>
1 rmse standard 2.09 5 0.265 Preprocessor1_Model1
2 rsq standard 0.673 5 0.0519 Preprocessor1_Model1
Tip
Set summarize = FALSE to see the individual
Summarize model fit CV metrics
Summarize model fit statistics from your CV iterations These statistics are calculated using the analysis set.
map_df(tips_fit_rs1$.extracts, ~ .x[[1]][[1]]) |>
summarise(mean_adj_rsq = mean(adj.r.squared),
mean_aic = mean(AIC),
mean_bic = mean(BIC))# A tibble: 1 × 3
mean_adj_rsq mean_aic mean_bic
<dbl> <dbl> <dbl>
1 0.670 434. 453.
Tip
Run the first line of code map_df(tips_fit_rs1$.extracts, ~ .x[[1]][[1]]) to see the individual
Another model - Model 2
Create the recipe for a new model that includes Party, Age, Meal, and Alcohol (an indicator for whether the party ordered alcohol with the meal). Conduct 10-fold cross validation and summarize the metrics.
Model 2: Recipe
# add code hereModel 2: Model building workflow
# add code hereModel 2: Conduct CV
Note
We will use the same folds as the ones used for Model 1. Why should we use the same folds to evaluate and compare both models?
# add code hereModel 2: Summarize assessment CV metrics
# add code hereModel 2: Summarize model fit CV metrics
# add code hereCompare and choose a model
Describe how the two models compare to each other based on cross validation metrics.
Which model do you choose for the final model? Why?
Fit the selected model
Fit the selected model using the entire training set.
# add code here
Tip
See notes for example code.
Evaluate the performance of the selected model on the testing data
Calculate predicted values
# add code hereCalculate
# add code here
Tip
See notes notes for example code.
How does the model performance on the testing data compare to its performance on the training data?
Is this what you expected? Why or why not?
Submission
Important
To submit the AE:
- Render the document to produce the PDF with all of your work from today’s class.
- Push all your work to your
ae-11repo on GitHub. (You do not submit AEs on Gradescope).