SLOS 2.0.0
Breaking changes
SLOS() now takes icu_column and
outcome (and an optional model) explicitly:
SLOS(data, icu_column, outcome, model = NULL).
load_SLOSModel() is replaced by
slos_load_pretrained_model() and
predict_and_evaluate() by
slos_predict_pretrained().
New features
- Complete workflow for building a new model from local data:
slos_pre_processing() (near-zero variance, correlation and
Cramér’s V screening, train/test split and multiple imputation with ),
slos_feature_select_rfe() and
slos_apply_feature_selection() (recursive feature
elimination), and slos_train_new_model() (stacked linear
model + random forest).
SLOS() accepts a model returned by
slos_train_new_model().
- New
show_funnel argument in SLOS().
- New vignette, “Getting started with SLOS”, and example scripts in
inst/examples.
Other changes
- The pre-trained model is downloaded to a temporary file, cached for
the R session, and a clear error is raised when there is no internet
connection or the download fails.
- Package documentation, examples and tests were rewritten.
SLOS 1.0.1