The goal of glmtree is to build decision trees with logistic regressions at their leaves, so that the resulting model mixes non parametric VS parametric and stepwise VS linear approaches to have the best predictive results, yet maintaining interpretability.
This is the implementation of glmtree as described in Formalization and study of statistical problems in Credit Scoring, Ehrhardt A. (see manuscript or web article)
You can install the development version of glmtree from Github with:
# install.packages("devtools")
devtools::install_github("adimajo/glmtree", build_vignettes = TRUE)Or alternatively directly from CRAN:
install.packages("glmtree")The help pages of this package can be reached by typing help(glmtree) (or ?glmtree) once the package is installed and loaded.
For further instructions, go to vignettes > glmtree.Rmd or go through the vignette directly in RStudio by typing vignette('glmtree') once the package is installed and loaded.
Both the help pages and the vignette are available online as a Github page built with pkgdown.