1 - Introduction

Machine learning with tidymodels

Welcome!

Who are you?

• You can use the magrittr %>% or base R |> pipe

• You are familiar with functions from dplyr, tidyr, ggplot2

• You have exposure to basic statistical concepts

• You do not need intermediate or expert familiarity with modeling or ML

Who are tidymodels?

• Simon Couch
• Hannah Frick
• Emil Hvitfeldt
• Max Kuhn

Many thanks to Davis Vaughan, Julia Silge, David Robinson, Julie Jung, Alison Hill, and Desirée De Leon for their role in creating these materials!

Asking for help

🟪 “I’m stuck and need help!”

🟩 “I finished the exercise”

👀

Tentative plan for this workshop

• Today:

  • Your data budget
  • What makes a model
  • Evaluating models

• Tomorrow:

  • Feature engineering
  • Tuning hyperparameters
  • Racing methods
  • Iterative search methods

Introduce yourself to your neighbors 👋

Check Slack (#ml-ws-2023) for an RStudio Cloud link.

What is machine learning?

What is machine learning?

What is machine learning?

Your turn

How are statistics and machine learning related?

How are they similar? Different?

03:00

the “two cultures”

model first vs. data first

inference vs. prediction

What is tidymodels?

library(tidymodels)
#> ── Attaching packages ──────────────────────────── tidymodels 1.1.0 ──
#> ✔ broom        1.0.5          ✔ rsample      1.1.1.9000
#> ✔ dials        1.2.0          ✔ tibble       3.2.1     
#> ✔ dplyr        1.1.2          ✔ tidyr        1.3.0     
#> ✔ infer        1.0.4          ✔ tune         1.1.1.9001
#> ✔ modeldata    1.1.0          ✔ workflows    1.1.3     
#> ✔ parsnip      1.1.0.9003     ✔ workflowsets 1.0.1     
#> ✔ purrr        1.0.1          ✔ yardstick    1.2.0.9001
#> ✔ recipes      1.0.6
#> ── Conflicts ─────────────────────────────── tidymodels_conflicts() ──
#> ✖ purrr::discard() masks scales::discard()
#> ✖ dplyr::filter()  masks stats::filter()
#> ✖ dplyr::lag()     masks stats::lag()
#> ✖ recipes::step()  masks stats::step()
#> • Use tidymodels_prefer() to resolve common conflicts.

The whole game

Part of any modelling process is

• Splitting your data into training and test set
• Using a resampling scheme
• Fitting models
• Assessing performance
• Choosing a model
• Fitting and assessing the final model

The whole game

The whole game

Stress that we are not fitting a model on the entire training set other than for illustrative purposes in deck 2.

The whole game

The whole game

The whole game

The whole game

The whole game

Let’s install some packages

If you are using your own laptop instead of RStudio Cloud:

install.packages("pak")

pkgs <- c("bonsai", "doParallel", "embed", "finetune", "lightgbm", "lme4", 
          "parallelly", "plumber", "probably", "ranger", "rpart", "rpart.plot", 
          "stacks", "textrecipes", "tidymodels", "tidymodels/modeldatatoo", 
          "vetiver")
pak::pak(pkgs)

Check Slack (#ml-ws-2023) for an RStudio Cloud link.

Our versions

bonsai (0.2.1.9000, Github (tidymodels/bonsai@aab79), broom (1.0.5, local), dials (1.2.0, CRAN), doParallel (1.0.17, CRAN), dplyr (1.1.2, CRAN), embed (1.0.0, CRAN), finetune (1.1.0.9000, Github (tidymodels/finetune@52d), ggplot2 (3.4.2, CRAN), lightgbm (3.3.5, CRAN), lme4 (1.1-33, CRAN), modeldata (1.1.0, CRAN), modeldatatoo (0.1.0.9000, Github (tidymodels/modeldatatoo), parallelly (1.36.0, CRAN), parsnip (1.1.0.9003, Github (tidymodels/parsnip@e627), plumber (1.2.1, CRAN), probably (1.0.2, CRAN), purrr (1.0.1, CRAN), ranger (0.15.1, CRAN), recipes (1.0.6, CRAN), rpart (4.1.19, CRAN), rpart.plot (3.1.1, CRAN), rsample (1.1.1.9000, Github (tidymodels/rsample@afc4), scales (1.2.1, CRAN), stacks (1.0.2.9000, local), textrecipes (1.0.2, CRAN), tibble (3.2.1, CRAN), tidymodels (1.1.0, CRAN), tidyr (1.3.0, CRAN), tune (1.1.1.9001, Github (tidymodels/tune@fea8b02), vetiver (0.2.0, CRAN), workflows (1.1.3, CRAN), workflowsets (1.0.1, CRAN), yardstick (1.2.0.9001, Github (tidymodels/yardstick@6c), and Quarto (1.3.433)