yardstick is an R package for estimating model performance using tidy data principles. It provides a consistent, dplyr-like syntax for calculating accuracy metrics on classification and regression models.
The package supports both binary and multiclass classification metrics with multiple estimation methods (macro, micro, hand-till). It works seamlessly with grouped data frames for calculating metrics across resamples, and includes autoplot methods for visualizing performance curves like ROC, precision-recall, and gain curves. All metrics return results in a consistent tibble format that integrates naturally with tidymodels workflows.
