MachineShop 3.1.0

Version Updates

3.1.0

• Add argument na.rm to dependence().
• Add global setting stats.VarImp for summary statistics to compute on permutation-based variable importance.
• Add permutation-based variable importance to varimp().
• Sort variable importance by first column only if not scaled.
• Correct the estimated variances for cross-validation estimators of mean performance difference in t.test.PerformanceDiff().
• Rename argument metric to type in varimp() functions for BartMachineModel, C50Model, EarthModel, RFSRCModel, and XGBModel.
• Set argument type default to "nsubsets" in EarthModel varimp().
• Expand case weighted metrics support.
  • Fix weights used in survival event-specific metrics.
  • Use weights for cross_entropy() numeric method.
  • Use weights for predicted survival probabilities.
• Fix error with argument f in roc_index() Surv method.

MachineShop 3.0.0

Version Updates

3.0.0

• Add slot weights to MLModel classes.
• Allow case weights in LMModel for all response types.
• Exclude infinite values from calculation of breaks in calibration().
• Fix invalid max = Inf arguments to print.default().
• Add support for case weights in performance metrics and curves.
• Evaluate ModelFrame() arguments strata and weights in data environment.
• Fix issue introduced in package version 2.9.0 of recipe case weights not being used in model fitting.
• Add column Weight of case weights to Resamples data frame.
• Rename values column to get_values in MLModel gridinfo slot.
• Move global settings resample_progress and resample_verbose to set_monitor() arguments progress and verbose.
• Move MLControl() arguments strata_breaks, strata_nunique, strata_prop, and strata_size to set_strata() arguments breaks, nunique, prop, and size.
• Move MLControl() arguments times, distr, and method to set_predict().
• Export %>% operator.
• Return case stratification values in the 'strata' slot of Resamples objects.