Introduction to tidymodels
What is tidymodels
tidymodels is a “meta-package” for modeling and statistical analysis that shares the underlying design philosophy, grammar, and data structures of the tidyverse.
Developed by
Max Kuhn
Julia Silge
When do we get to play with tidymodels
Installing tidymodels
When loading the package, the versions and conflicts are listed:
The ecosystem
More inconsistency 🤢
tidymodels Consistency 😎
Through the parsnip package we are provided with a time saving framework for exploring multiple models!!
Example:
# Logistic Regression
logistic_reg_glm_spec <-
logistic_reg() %>%
set_engine('glm') %>%
set_mode('classification')
# Decision Tree
decision_tree_rpart_spec <-
decision_tree(
tree_depth = tune(),
min_n = tune(),
cost_complexity = tune()
) %>%
set_engine("rpart") %>%
set_mode("classification")
# Bagged MARS Model
bag_mars_earth_spec <-
bag_mars() %>%
set_engine('earth') %>%
set_mode('classification')
# Naive Bayes
naive_Bayes_naivebayes_spec <-
naive_Bayes(smoothness = tune(), Laplace = tune()) %>%
set_engine('naivebayes') %>%
set_mode('classification')
# Random Forest
rand_forest_randomForest_spec <-
rand_forest(mtry = tune(), min_n = tune()) %>%
set_engine('randomForest') %>%
set_mode('classification')
If I haven’t convinced you yet
The real power of tidymodels is baked into the recipes package.
Binds a sequence of preprocessing steps to a training data set.
Defines the roles that the variables are to play in the design matrix.
Specifies what data cleaning needs to take place, and what feature engineering needs to happen.
Recap
- We know what
tidymodelsis - We understand its importance
- Lets starts coding…
Data
2019 Youth Risk Behavioral Surveillance System
EDA
Let’s explore the relationship between difficulty concentrating, diet, sleep, physical activity and hours of playing video games.
Tidymodels
Building Blocks tidymodels
Recipes
- Every model requires a design matrix as an input.
- Design Matrix: tidy data, with one obervation per row and one predictor per column.
HOWEVER
Design matrices do not always come in the required format:
- KNN needs normalized predictors
- A linear model requires categorical predictors to be one-hot encoded
- Logistic regression needs complete data (imputation)
Baking the Recipe – Declarative Programming
This step is crucial!
You have to check your data after the recipe to make sure the transformations look alright.
# A tibble: 13,677 × 24
DrinkFruitJuice EatFruit EatSalad EatPotatoes EatCarrots EatOtherVeggies
<dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 2 3 1 1 1 2
2 2 7 6 5 5 6
3 1 4 2 2 2 5
4 2 2 1 2 2 3
5 4 5 1 2 3 4
6 2 1 2 1 1 2
7 1 4 1 2 1 2
8 4 2 2 1 2 3
9 2 3 3 1 1 3
10 2 2 1 1 1 3
# ℹ 13,667 more rows
# ℹ 18 more variables: DrinkSoda <dbl>, DrinkMilk <dbl>, EatBreakfast <dbl>,
# PhysicalActivity <dbl>, HoursTV <dbl>, HoursVideoGames <dbl>,
# HoursSleep <dbl>, SportsDrinks <dbl>, DrinksWater <dbl>,
# ConcussionSports <dbl>, DifficultyConcentrating <fct>, Sex_Male <dbl>,
# Grade_X11 <dbl>, Grade_X12 <dbl>, Grade_X9 <dbl>,
# SexOrientation_Gay.or.Lesbian <dbl>, SexOrientation_Heterosexual <dbl>, …Recipes in ONE image
Parsnip
A model specification has three individual components:
- Type: The model type that is about to be fitted (e.g., linear/logit regression, random forest or SVM).
- Mode: The mode of prediction, i.e. regression or classification.
- Engine: The computational engine implemented in
Rwhich usually corresponds to a certain modeling function (lm,glm), package (e.g.,rpart,glmnet,randomForest) or computing framework (e.g.,Stan,sparklyr).
Setting the Specifications ![]()
Logistic Regression Model Specification (classification)
Computational engine: glm