The ingredients
package is a collection of tools for assessment of feature importance and feature effects.
Key functions:
feature_importance()
for assessment of global level feature importance,ceteris_paribus()
for calculation of the Ceteris Paribus / What-If Profiles (read more at https://pbiecek.github.io/PM_VEE/ceterisParibus.html),partial_dependency()
for Partial Dependency Plots,conditional_dependency()
for Conditional Dependency Plots also called M Plots,accumulated_dependency()
for Accumulated Local Effects Plots,aggregate_profiles()
andcluster_profiles()
for aggregation of Ceteris Paribus Profiles,aspect_importance()
for LIME style explanations,calculate_oscillations()
for calculation of the Ceteris Paribus Oscillations (read more at https://pbiecek.github.io/PM_VEE/ceterisParibusOscillations.html),ceteris_paribus_2d()
for Ceteris Paribus 2D Profiles (read more at https://pbiecek.github.io/PM_VEE/ceterisParibus2d.html),- generic
print()
andplot()
for better usability of selected explainers, - generic
plotD3()
for interactive, D3 based explanations, - generic
describe()
for explanations in natural language.
The philosophy behind ingredients
explanations is described in the Predictive Models: Explore, Explain, and Debug e-book. The ingredients
package is a part of DrWhy.AI universe.
# the easiest way to get ingredients is to install it from CRAN:
install.packages("ingredients")
# Or the the development version from GitHub:
# install.packages("devtools")
devtools::install_github("ModelOriented/ingredients")
feature_importance()
, ceteris_paribus()
and aggregated_profiles()
also work with D3!
see an example
Work on this package was financially supported by the 'NCN Opus grant 2016/21/B/ST6/02176'.