SchlossLab/mikropml

New vignette for visualizing performance and feature importance

Opened this issue · 0 comments

kelly-sovacool commented

Currently, a lot of plots are buried in the parallel vignette. We should break them out into a separate vignette.

  • plots for a single model
    • ROC and PRC
    • Feature importance with 95% CI
  • plots for multiple models on same dataset
    • Mean ROC and PRC
    • Performance boxplots with AUROC and AUPRC
    • Feature importance with 95% CI
  • plots for model(s) on different datasets with class imbalances
    • Mean balanced PRC
    • Boxplot with AUBPRC