A LearnerSurv for a survival random forest implemented in randomForestSRC::rfsrc() in package randomForestSRC.

Format

R6::R6Class() inheriting from LearnerSurv.

Details

rfsrc has three prediction outcomes, from the fitted model these are respectively:

  1. predicted - This is ensemble mortality defined in Ishwaran et al. (2008), the sum of an individuals cumulative hazard function over all time-points

  2. chf - Cumulative hazard function, estimated via a bootstrapped Nelson-Aalen estimator (Ishwaran, 2008)

  3. surv - Survival function, estimated via a bootrstrapped Kaplan-Meier estimate (https://kogalur.github.io/randomForestSRC/theory.html)

Only the second two of these are returned in the distr predict.type, as Nelson-Aalen and Kaplan-Meier will give different results, the estimator can be chosen in the parameter set under "estimator".

The 'risk' predict.type is here defined as the mean of the cumulative hazard over all unique death times.

Construction

LearnerSurvRandomForestSRC$new()
mlr_learners$get("surv.randomForestSRC")
lrn("surv.randomForestSRC")

References

Ishwaran H. and Kogalur U.B. (2019). Fast Unified Random Forests for Survival, Regression, and Classification (RF-SRC), R package version 2.9.1.

Ishwaran H., Kogalur U.B., Blackstone E.H. and Lauer M.S. (2008). Random survival forests. Ann. Appl. Statist. 2(3), 841--860.

Breiman, L. (2001). Random Forests. Machine Learning 45(1). doi: 10.1023/A:1010933404324 .

See also