RSF: RSF: Random Survival Forest

In nguforche/MLSurvival: Machine Learning for Survival Analysis

Description

Train the random survival forest through the ranger package. The optimal RSF tuning parameters: min.node.size,mtry, and splitrule can be selected through grid search.

Usage

train_RSF(form, dat, newdata = NULL, predict.times, trControl = NULL,
  seed = 123, parallel = FALSE, mc.cores = 2, ...)

Arguments

form	survival formula
dat	data frame
predict.times	survival prediction times
trControl	list of control parameters: ntrees: mumber of trees number: number of cross-validations tuneLength: tuning paramer grid size importance: ranger variable importance
parallel	run cross-validation in parallel? Uses mclapply which works only on linux
...	further arguments passed to caret or other methods.
tuneLength	same as tuneLength in the caret package

Value

returns a list with items:

• finalModel: final model trained on the complete data (dat) using optimal tuning paramters

• fitted: predictions on complete data (dat)

• threshold: optimal classification threshold

• resamples: cross-validation results: predictions on resampled data

• predict.times: survival prediction times

• bestTune: optimal tuning parameters

nguforche/MLSurvival documentation built on July 28, 2019, 1:59 p.m.