tidymodels-overview

The tidymodels ecosystem is a collection of R packages for end-to-end machine learning workflows. Core packages like recipes, parsnip, workflows, tune, rsample, yardstick, workflowsets, and stacks map directly to pipeline steps from data splitting to model evaluation.

To plan an end-to-end machine learning workflow in R, use tidymodels packages sequentially: rsample for data splitting, recipes for preprocessing, parsnip for modeling, workflows for bundling, tune for hyperparameters, and yardstick for evaluation.

Yes, you can use tidymodels to manage both data preparation and model evaluation. The recipes package handles data preparation and feature engineering, while yardstick calculates evaluation metrics, all connected through the workflows package.

The best way to structure an ML project in R is to map the pipeline using the tidymodels ecosystem overview. Understand the relationships between rsample, recipes, parsnip, workflows, tune, and stacks to outline steps from data splitting to model ensembling.

Use workflowsets to compare multiple preprocessing and model combinations simultaneously, and use stacks to ensemble multiple tidymodels models together. Both are applied during the tuning and evaluation phases of an end-to-end R machine learning workflow.