probably

probably is an R package for working with predicted probabilities from classification and regression models. It converts probabilities to class predictions, identifies optimal probability thresholds, and handles cases where predictions are too uncertain to be reliable.

The package provides tools for model calibration assessment and correction, lets you define equivocal zones where probability predictions are too ambiguous to make a definitive call, and integrates with the tidymodels ecosystem for threshold optimization. It addresses the common problem that raw model probabilities often need post-processing before making final predictions, especially when uncertainty varies across the prediction space.