Development in November 2019.
- Added a tuner for
ds_rf()
that picks optimalmtry
values.
Development in October 2019.
reg_logreg()
andds_reg_logreg()
implement a self-tuning regularized logistic regression model based on the glmnet package. Tuning is performed via cross-validation and successively picks alpha and then lambda values.rf()
andds_rf()
implement a random forest probability tree model using the ranger package.
Initial version in September 2019. This had the following models:
logistic_reg()
is a standard logistic regression model, withds_logistic_reg()
as a wrapper for modeling democratic spaces. It includes an option to standardize features prior to model estimation ("normalize" argument), using a standardizer function made bymake_standardizer()
.logistic_reg_featx()
andds_logistic_reg_featx()
are standard logistic regression models with a feature extraction pre-processing step for the input feature data. This uses PCA (the only method implemented currently) to reduce the number of numeric input features to 5, viamake_extract_features()
.
The GitHub repo, but not package, includes a template for adding new models in add_new_model.R
.