In this chapter, you’ll explore the rich ecosystem of R packages that power tidymodels and learn how they can streamline your machine learning workflows. You’ll then put your tidymodels skills to the test by predicting house sale prices in Seattle, Washington.

Het tidymodels-ecosysteem

Opties

Data-resampling en feature engineering

Model fitten en afstemmen

Model evalueren

Tidymodels-pakketten

Trainings- en testgegevenssets maken

Verdeling van waarden van de uitkomstvariabele

Lineaire regressie met tidymodels

Een lineair regressiemodel fitten

Geschatte modelparameters verkennen

Verkoopprijzen van huizen voorspellen

Modelprestaties evalueren

Modelprestaties meten

R-squared-plot

Volledig model-fitproces met last_fit()

Machine Learning with tidymodels

Learn how to predict categorical outcomes by training classification models. Using the skills you’ve gained so far, you’ll predict the likelihood of customers canceling their service with a telecommunications company.

Classification models

Data resampling

Fitting a logistic regression model

Combining test dataset results

Assessing model fit

Calculating metrics from the confusion matrix

Evaluating performance with yardstick

Creating custom metric sets

Visualizing model performance

Plotting the confusion matrix

ROC curves and area under the ROC curve

Automating the modeling workflow

Streamlining the modeling process

Collecting predictions and creating custom metrics

Complete modeling workflow

Classification Models

Find out how to bake feature engineering pipelines with the recipes package. You’ll prepare numeric and categorical data to help machine learning algorithms optimize your predictions.

Feature engineering

Exploring recipe objects

Creating recipe objects

Training a recipe object

Numeric predictors

Discovering correlated predictors

Removing correlated predictors with recipes

Multiple feature engineering steps

Nominal predictors

Applying step_dummy() to predictors

Ordering of step_*() functions

Complete feature engineering pipeline

Feature engineering process

Model training and prediction

Model performance metrics

Feature Engineering

Now it’s time to streamline the modeling process using workflows and fine-tune models with cross-validation and hyperparameter tuning. You’ll learn how to tune a decision tree classification model to predict whether a bank's customers are likely to default on their loan.

Machine learning workflows

Exploring the loans dataset

Specifying a model and recipe

Creating workflows

Estimating performance with cross validation

Measuring performance with cross validation

Cross validation with logistic regression

Comparing model performance profiles

Hyperparameter tuning

Setting model hyperparameters

Random grid search

Exploring tuning results

Selecting the best model

Finalizing a workflow

Training a finalized workflow

Congratulations!

Workflows and Hyperparameter Tuning

Home sales

Loans

Telecom data

Employee data

Purchase leads

Tidymodels is een krachtige verzameling R-pakketten die Machine Learning-workflows vereenvoudigt. Leer gegevenssets te splitsen voor cross-validatie, data voor te bewerken met het recipes-pakket van tidymodels en Machine Learning-algoritmen te finetunen. Je leert kernconcepten zoals het definiëren van modelobjecten en het opzetten van modelworkflows. Daarna pas je je vaardigheden toe om huizenprijzen te voorspellen en medewerkers te classificeren op hun risico om een bedrijf te verlaten.

Het tidymodels-ecosysteem

Create Your Free Account