Introducing tidymodels

Machine learning best practices with tidymodels

The tidymodels family of R packages provides a framework for predictive modeling and machine learning following the principles of the tidyverse family of R packages.

In addition to augmenting tidyverse with machine learning capabilities, tidymodels has been designed to promote machine learning best practices in R code. For example, it helps combat information leakage (a concept covered later in the course) in machine learning work.

This course will demonstrate the use of the following packages to develop machine learning code:

• rsample: Is used early in the machine learning project lifecycle to split data into training and test sets.

• recipes: Enables building data preparation pipelines step by step because data needs to be prepared before machine learning algorithms can effectively use it.

• parsnip: Provides a unified interface for specifying the training of machine learning models. For example, specifying a CART classification tree as the model to be trained.

• dials: Provides functions that help tune machine learning models. For example, by creating and using tuning grids.

• tune: Helps optimize the hyperparameter settings of machine learning algorithms.

• yardstick: Provides functions for measuring the effectiveness of machine learning model predictions using performance metrics like accuracy, sensitivity, and specificity.

• workflows: Provides functions that allow for orchestrating all the functionality in this list.