| cheem {cheem} | R Documentation |
cheem
Description
Consider an arbitrary black-box model. Local explanations are a class of
point-measures of the variable importance to the model for one location in
the explanatory variables. The package cheem extracts the local attribution
of all observations and first creates a 2D approximated space comparing
data- and attribution-spaces. After a primary and comparison observation have
been selected the attribution of the primary is used as the 1D basis for a
manual tour. This tour changes the contribution from the variable that
deviates the most from its expected value. By viewing the positions of the
primary and comparison points, the analyst can scrutinize the explanation
identify variable attribution leading to a misclassification or large
residual.
Details
GitHub: https://github.com/nspyrison/cheem
Author(s)
Maintainer: Nicholas Spyrison spyrison@gmail.com (ORCID)
See Also
cheem_ls() or run_app() for help getting started