New function: “rconfusionmatrix” - allows for a new way ofcalculating weighted accuracy metrics, by reallocating observationsclassified in the vicinity of the true category to the confusion matrixdiagonal.
Bugfix: The “wconfusionmatrix” weighting calculation is adjusted tofix a rare bug which would appear when selecting a specific set ofweights.
Improvements: Vignette updated to include additional examples andimprovements in formatting and display of results.
New function: “balancedaccuracy” - implements the Starovoitov-Golubsine-based accuracy metrics, specifically designed for classificationwith imbalanced data.
New features: the “wconfusionmatrix” and “weightmatrix” functions nowinclude two additional weighting schemes “sin” and “tanh” - based on thesinus and hyperbolic tangent functions, which allow for even greatercustomization of weights assigned to classes.
Bugfix: The “interval” weighting scheme under the “wconfusionmatrix”and “weightmatrix” functions has been fixed. It previously producedinaccurate results when run with matrix dimensions greater than 5.
Improvements: Vignette updated to include additional examples andimprovements in formatting and display of results.
This is the initial release.