Model Agnostic Explainers for Individual Predictions

Model agnostic tool for decomposition of predictions from black boxes. Break Down Table shows contributions of every variable to a final prediction. Break Down Plot presents variable contributions in a concise graphical way. This package work for binary classifiers and general regression models.


breakDown 0.1.6

  • broken.default has now the keep_distributions arguments. If TRUE then the whole distribution of conditional residuals is remebered and avaliable for plotting #17
  • small updates in

breakDown 0.1.5

  • small changes in broken.default to make it work with xgboost and other non data.frame data
  • broken.lm supports unnormalized coefficients (thanks to Joseph Larmarange) just add predict.function = betas #9

breakDown 0.1.4

  • broken.default is now model agnostic!
  • broken.ranger is removed since broken.default is much better
  • small fixes in print and plot functions, a new vigniette for model agnostic plots

breakDown 0.1.3

  • small fixes and submission to CRAN

breakDown 0.1.2

  • broken.lm and broken.glm are now supporting interactions (#7)
  • print() and plot() functions are now handling different options for rounding via additional arguments digits = 3, rounding_function = round (#8)

breakDown 0.1.1

  • the baseline argument is added to the broken function (#1)
  • vignettes for lm and glm models are added (#2)

breakDown 0.1

  • waterfall like plots and support for lm models
  • waterfall like plots and support for glm models
  • HR dataset added

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


0.2.1 by Przemyslaw Biecek, 8 months ago

Report a bug at

Browse source code at

Authors: Przemyslaw Biecek [aut, cre] , Aleksandra Grudziaz [ctb]

Documentation:   PDF Manual  

GPL-2 license

Imports ggplot2

Suggests knitr, rmarkdown, e1071, kernlab, xgboost, caret, randomForest, DALEX, ranger, testthat

Imported by live, modelDown, survxai.

Suggested by xspliner.

See at CRAN