pdp
Overview
pdp is an R package for constructing partial dependence plots (PDPs) and individual conditional expectation (ICE) curves. PDPs and ICE curves are part of a larger framework referred to as interpretable machine learning (IML), which also includes (but not limited to) variable importance plots (VIPs). While VIPs (available in the R package vip) help visualize feature impact (either locally or globally), PDPs and ICE curves help visualize feature effects. An in-progress, but comprehensive, overview of IML can be found at the following URL: https://github.com/christophM/interpretable-ml-book.
A detailed introduction to pdp has been published in The R Journal: “pdp: An R Package for Constructing Partial Dependence Plots”, https://journal.r-project.org/archive/2017/RJ-2017-016/index.html. You can track development at https://github.com/bgreenwell/pdp. To report bugs or issues, contact the main author directly or submit them to https://github.com/bgreenwell/pdp/issues. For additional documentation and examples, visit the package website.
As of right now, pdp
exports the following functions:
partial()
- compute partial dependence functions and individual conditional expectations (i.e., objects of class"partial"
and"ice"
, respectively) from various fitted model objects;plotPartial()"
- constructlattice
-based PDPs and ICE curves;autoplot()
- constructggplot2
-based PDPs and ICE curves;see vip instead for a more robust and flexible replacement;topPredictors()
extract most “important” predictors from various types of fitted models.exemplar()
- construct an exemplar record from a data frame (experimental feature that may be useful for constructing fast, approximate feature effect plots.)
Installation
# The easiest way to get pdp is to install it from CRAN:
install.packages("pdp")
# Alternatively, you can install the development version from GitHub:
if (!("remotes" %in% installed.packages()[, "Package"])) {
install.packages("remotes")
}
remotes::install_github("bgreenwell/pdp")