use a "one-vs-all" SMOTE subsampling technique: over-sample each class vs. the other classes and combine the oversampled datasets to create the "balanced" dataset
new multiclass metric G-mean
increase minimum R version to 3.6.0
move packages used for classification to Suggests to reduce the number of dependencies and are used conditionally
remove deprecated context() from tests
put the macro and micro averaged ROC curves at the end of legend in roc_plot()
suppress warnings in call to multiROC::multi_roc() after updates to stats:::regularize.values() in R-3.6.0 passes warn.collapsing = TRUE if there is no value for ties in stats::approx()
use a "one-vs-all" SMOTE subsampling technique: over-sample each class vs. the other classes and combine the oversampled datasets to create the "balanced" dataset
new multiclass metric G-mean
increase minimum R version to 3.6.0
move packages used for classification to
Suggeststo reduce the number of dependencies and are used conditionallyremove deprecated
context()from testsput the macro and micro averaged ROC curves at the end of legend in
roc_plot()suppress warnings in call to
multiROC::multi_roc()after updates tostats:::regularize.values()in R-3.6.0 passeswarn.collapsing = TRUEif there is no value fortiesinstats::approx()