dblr - Discrete Boosting Logistic Regression
Trains logistic regression model by discretizing
continuous variables via gradient boosting approach. The
proposed method tries to achieve a tradeoff between
interpretation and prediction accuracy for logistic regression
by discretizing the continuous variables. The variable binning
is accomplished in a supervised fashion. The model trained by
this package is still a single logistic regression model, but
not a sequence of logistic regression models. The fitted model
object returned from the model training consists of two tables.
One table is used to give the boundaries of bins for each
continuous variable as well as the corresponding coefficients,
and the other one is used for discrete variables. This package
can also be used for binning continuous variables for other
statistical analysis.
