Data Science with Tidymodels, Workflows and Recipes

How to use Tidymodels, Workflows and Recipes for Regression
Data Science
Statistics
Data preprocessing
Regression
R
Autor:in

Jan Kirenz

Veröffentlichungsdatum

19. Dezember 2019

Geändert

25. Dezember 2023

Data Science project with Regression

The following examples are adapted from the excellent book “Hands-on machine learning with scikit-learn, keras and tensorflow” from A. Geron and the tidymodels documentation

In this tutorial you will learn how to specify a simple regression model with the tidymodels package using recipes, which is designed to help you preprocess your data before training your model.

To use the code in this article, you will need to install the following packages:

If you like to install all packages at once, use the code below.

install.packages(c("tidyverse", "tidymodels", "skimr", "GGally", "ggmap", "kableExtra")) 
library(tidyverse)
library(tidymodels)
library(skimr)
library(GGally)
library(ggmap)
library(kableExtra)

In this example, our goal is to build a model of housing prices in California. In particular, the model should learn from California census data and be able to predict the median house price in any district (population of 600 to 3000 people), given some predictor variables. We use the root mean square error (RMSE) as a performance measure for our regression problem.

Data understanding

In Data Understanding, we first

  • Import data
  • Get an overview about the data structure
  • Discover and visualize the data to gain insights

Import Data

First of all, let’s import the data:

LINK <- "https://raw.githubusercontent.com/kirenz/datasets/master/housing.csv"
housing_df <- read_csv(LINK)

Data overview

Next, we take a look at the data structure:

California census top 4 rows of the DataFrame:

head(housing_df, 4)
# A tibble: 4 × 10
  longitude latitude housing_median_age total_rooms total_bedrooms population
      <dbl>    <dbl>              <dbl>       <dbl>          <dbl>      <dbl>
1     -122.     37.9                 41         880            129        322
2     -122.     37.9                 21        7099           1106       2401
3     -122.     37.8                 52        1467            190        496
4     -122.     37.8                 52        1274            235        558
# ℹ 4 more variables: households <dbl>, median_income <dbl>,
#   median_house_value <dbl>, ocean_proximity <chr>

Data info:

glimpse(housing_df)
Rows: 20,640
Columns: 10
$ longitude          <dbl> -122.23, -122.22, -122.24, -122.25, -122.25, -122.2…
$ latitude           <dbl> 37.88, 37.86, 37.85, 37.85, 37.85, 37.85, 37.84, 37…
$ housing_median_age <dbl> 41, 21, 52, 52, 52, 52, 52, 52, 42, 52, 52, 52, 52,…
$ total_rooms        <dbl> 880, 7099, 1467, 1274, 1627, 919, 2535, 3104, 2555,…
$ total_bedrooms     <dbl> 129, 1106, 190, 235, 280, 213, 489, 687, 665, 707, …
$ population         <dbl> 322, 2401, 496, 558, 565, 413, 1094, 1157, 1206, 15…
$ households         <dbl> 126, 1138, 177, 219, 259, 193, 514, 647, 595, 714, …
$ median_income      <dbl> 8.3252, 8.3014, 7.2574, 5.6431, 3.8462, 4.0368, 3.6…
$ median_house_value <dbl> 452600, 358500, 352100, 341300, 342200, 269700, 299…
$ ocean_proximity    <chr> "NEAR BAY", "NEAR BAY", "NEAR BAY", "NEAR BAY", "NE…

Data summary of numerical and categorical attributes using a function from the package skimr:

skim(housing_df)
Data summary
Name housing_df
Number of rows 20640
Number of columns 10
_______________________
Column type frequency:
character 1
numeric 9
________________________
Group variables None

Variable type: character

skim_variable n_missing complete_rate min max empty n_unique whitespace
ocean_proximity 0 1 6 10 0 5 0

Variable type: numeric

skim_variable n_missing complete_rate mean sd p0 p25 p50 p75 p100 hist
longitude 0 1.00 -119.57 2.00 -124.35 -121.80 -118.49 -118.01 -114.31 ▂▆▃▇▁
latitude 0 1.00 35.63 2.14 32.54 33.93 34.26 37.71 41.95 ▇▁▅▂▁
housing_median_age 0 1.00 28.64 12.59 1.00 18.00 29.00 37.00 52.00 ▃▇▇▇▅
total_rooms 0 1.00 2635.76 2181.62 2.00 1447.75 2127.00 3148.00 39320.00 ▇▁▁▁▁
total_bedrooms 207 0.99 537.87 421.39 1.00 296.00 435.00 647.00 6445.00 ▇▁▁▁▁
population 0 1.00 1425.48 1132.46 3.00 787.00 1166.00 1725.00 35682.00 ▇▁▁▁▁
households 0 1.00 499.54 382.33 1.00 280.00 409.00 605.00 6082.00 ▇▁▁▁▁
median_income 0 1.00 3.87 1.90 0.50 2.56 3.53 4.74 15.00 ▇▇▁▁▁
median_house_value 0 1.00 206855.82 115395.62 14999.00 119600.00 179700.00 264725.00 500001.00 ▅▇▅▂▂

Count levels of our categorical variable:

housing_df |> 
  count(ocean_proximity,
        sort = TRUE)
# A tibble: 5 × 2
  ocean_proximity     n
  <chr>           <int>
1 <1H OCEAN        9136
2 INLAND           6551
3 NEAR OCEAN       2658
4 NEAR BAY         2290
5 ISLAND              5

The function ggscatmat from the package GGally creates a matrix with scatterplots, densities and correlations for numeric columns. In our code, we enter the dataset housing_df, choose columns 6 to 9, a color column for our categorical variable ocean_proximity, and an alpha level of 0.8 (for transparency).

ggscatmat(housing_df, columns = 6:9, color="ocean_proximity", alpha=0.8)

To obtain an overview of even more visualizations, we can use the function ggpairs:

ggpairs(housing_df)

Data exploration

A Geographical scatterplot of the data:

housing_df |> 
  ggplot(aes(x = longitude, y = latitude)) +
  geom_point(color = "cornflowerblue")

Scatterplot of longitude and latitude

A better visualization that highlights high-density areas:

housing_df |> 
  ggplot(aes(x = longitude, y = latitude)) +
  geom_point(color = "cornflowerblue", alpha = 0.1) 

Scatterplot of longitude and latitude that highlights high-density areas

California housing prices:

  • red is expensive,
  • purple is cheap and
  • larger circles indicate areas with a larger population.
housing_df |> 
  ggplot(aes(x = longitude, y = latitude)) +
  geom_point(aes(size = population, color = median_house_value), 
             alpha = 0.4) +
  scale_colour_gradientn(colours=rev(rainbow(4)))

California housing_df prices

Next, you need to creata a stadiamaps account to display they map.

ggmap::register_stadiamaps("enter-your-AP")
library(ggmap)

qmplot(x = longitude, 
       y = latitude, 
       data = housing_df, 
       geom = "point", 
       color = median_house_value, 
       size = population,
       alpha = 0.4) +
  scale_colour_gradientn(colours=rev(rainbow(4)))

Data preparation

Data splitting

Before we build our model, we first split data into training and test set using stratified sampling.

Let’s assume we would know that the median income is a very important attribute to predict median housing prices. Therefore, we would want to create a training and test set using stratified sampling.

A stratum (plural strata) refers to a subset (part) of the population (entire collection of items under consideration) which is being sampled:

housing_df |> 
  ggplot(aes(median_income)) +
  geom_histogram(bins = 30)

Histogram of Median Income

We want to ensure that the test set is representative of the various categories of incomes in the whole dataset. In other words, we would like to have instances for each stratum, or else the estimate of a stratum’s importance may be biased. This means that you should not have too many strata, and each stratum should be large enough. We use 5 strata in our example.

set.seed(42)

new_split <- initial_split(housing_df, 
                           prop = 3/4, 
                           strata = median_income, 
                           breaks = 5)

new_train <- training(new_split) 
new_test <- testing(new_split)

Recipes

Next, we use a recipe() to build a set of steps for data preprocessing and feature engineering.

Recipes are built as a series of preprocessing steps, such as:

  • converting qualitative predictors to indicator variables (also known as dummy variables),
  • transforming data to be on a different scale (e.g., taking the logarithm of a variable),
  • transforming whole groups of predictors together,
  • extracting key features from raw variables (e.g., getting the day of the week out of a date variable),

In summary, the idea of the recipes package is to define a recipe or blueprint that can be used to sequentially define the encodings and preprocessing of the data (i.e. “feature engineering”) before we build our models.

  1. First, we must tell the recipe() what our model is going to be (using a formula here) and what our training data is.

  2. step_novel() will convert all nominal variables to factors.

  3. We then convert the factor columns into (one or more) numeric binary (0 and 1) variables for the levels of the training data.

  4. We remove any numeric variables that have zero variance.

  5. We normalize (center and scale) the numeric variables.

housing_rec <-
  recipe(median_house_value ~ ., data = new_train) |>
  step_novel(all_nominal(), -all_outcomes()) |>
  step_dummy(all_nominal()) |>
  step_zv(all_predictors()) |>
  step_normalize(all_predictors())

# Show the content of our recipe
housing_rec

Now it’s time to specify and then fit our models.

Model building

Model specification

  1. Pick a model type: choose from this list
  2. Set the engine: choose from this list
  3. Set the mode: regression or classification
library(tidymodels)

lm_spec <- # your model specification
  linear_reg() |>  # model type
  set_engine(engine = "lm") |>  # model engine
  set_mode("regression") # model mode

# Show your model specification
lm_spec
Linear Regression Model Specification (regression)

Computational engine: lm 

To combine the data preparation with the model building, we use the package workflows.

A workflow is an object that can bundle together your pre-processing, modeling, and post-processing requests

Create workflow

lm_wflow <-
 workflow() |>
 add_model(lm_spec) |> 
 add_recipe(housing_rec)

Evaluate model

We build a validation set with K-fold crossvalidation:

set.seed(100)

cv_folds <-
 vfold_cv(new_train, 
          v = 5, 
          strata = median_income,
          breaks = 5) 

cv_folds
#  5-fold cross-validation using stratification 
# A tibble: 5 × 2
  splits               id   
  <list>               <chr>
1 <split [12380/3099]> Fold1
2 <split [12382/3097]> Fold2
3 <split [12384/3095]> Fold3
4 <split [12385/3094]> Fold4
5 <split [12385/3094]> Fold5

Now we can fit the model and collect the performance metrics with collect_metrics():

lm_wflow_eval <- 
  lm_wflow |> 
  fit_resamples(
    median_house_value ~ ., 
    resamples = cv_folds
    ) 

lm_wflow_eval|> 
    collect_metrics()
# A tibble: 2 × 6
  .metric .estimator      mean     n   std_err .config             
  <chr>   <chr>          <dbl> <int>     <dbl> <chr>               
1 rmse    standard   69359.        5 549.      Preprocessor1_Model1
2 rsq     standard       0.643     5   0.00432 Preprocessor1_Model1

Usually, we would fit multiple models and select the one with the smallest RMSE. In this example, we only demonstrate the process with one model.

Last fit and evaluation

Fit the best model to the training set and evaluate the test set with the function last_fit():

last_fit_lm <- last_fit(lm_wflow, split = new_split)

# Show RMSE and RSQ
last_fit_lm |> 
  collect_metrics()
# A tibble: 2 × 4
  .metric .estimator .estimate .config             
  <chr>   <chr>          <dbl> <chr>               
1 rmse    standard   67113.    Preprocessor1_Model1
2 rsq     standard       0.651 Preprocessor1_Model1