Survival Gradient Boosting Machine Learner

Gradient Boosting for Survival Analysis. Calls gbm::gbm() from gbm.

Prediction types

This learner returns two prediction types, using the internal predict.gbm() function:

1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation.

2. crank: same as lp.

Dictionary

This Learner can be instantiated via lrn():

lrn("surv.gbm")

Meta Information

• Task type: “surv”

• Predict Types: “crank”, “lp”

• Feature Types: “integer”, “numeric”, “factor”, “ordered”

• Required Packages: mlr3, mlr3proba, mlr3extralearners, gbm

Parameters

Id	Type	Default	Levels	Range
distribution	character	coxph	coxph	-
n.trees	integer	100		\([1, \infty)\)
cv.folds	integer	0		\([0, \infty)\)
interaction.depth	integer	1		\([1, \infty)\)
n.minobsinnode	integer	10		\([1, \infty)\)
shrinkage	numeric	0.001		\([0, \infty)\)
bag.fraction	numeric	0.5		\([0, 1]\)
train.fraction	numeric	1		\([0, 1]\)
keep.data	logical	FALSE	TRUE, FALSE	-
verbose	logical	FALSE	TRUE, FALSE	-
var.monotone	untyped	-		-
n.cores	integer	1		\((-\infty, \infty)\)
single.tree	logical	FALSE	TRUE, FALSE	-

Initial parameter values

• distribution:

• Actual default: "bernoulli"

• Adjusted default: "coxph"

• Reason for change: This is the only distribution available for survival.

• keep.data:

  • Actual default: TRUE

  • Adjusted default: FALSE

  • Reason for change: keep.data = FALSE saves memory during model fitting.

• n.cores:

  • Actual default: NULL

  • Adjusted default: 1

  • Reason for change: Suppressing the automatic internal parallelization if cv.folds > 0 and avoid threading conflicts with future.

References

Friedman, H J (2002). “Stochastic gradient boosting.” Computational statistics & data analysis, 38(4), 367–378.

See also

• Dictionary of Learners: mlr3::mlr_learners.

• as.data.table(mlr_learners) for a table of available Learners in the running session (depending on the loaded packages).

• Chapter in the mlr3book: https://mlr3book.mlr-org.com/basics.html#learners

• mlr3learners for a selection of recommended learners.

• mlr3cluster for unsupervised clustering learners.

• mlr3pipelines to combine learners with pre- and postprocessing steps.

• mlr3tuning for tuning of hyperparameters, mlr3tuningspaces for established default tuning spaces.

Author

RaphaelS1

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvGBM

Methods

Public methods

• LearnerSurvGBM$new()

• LearnerSurvGBM$importance()

• LearnerSurvGBM$clone()

Inherited methods

• mlr3::Learner$base_learner()
• mlr3::Learner$encapsulate()
• mlr3::Learner$format()
• mlr3::Learner$help()
• mlr3::Learner$predict()
• mlr3::Learner$predict_newdata()
• mlr3::Learner$print()
• mlr3::Learner$reset()
• mlr3::Learner$train()

---

Method new()

Creates a new instance of this R6 class.

Usage

LearnerSurvGBM$new()

---

Method importance()

The importance scores are extracted from the model slot variable.importance.

Usage

LearnerSurvGBM$importance()

Returns

Named numeric().

---

Method clone()

The objects of this class are cloneable with this method.

Usage

LearnerSurvGBM$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

# Define the Learner
learner = mlr3::lrn("surv.gbm")
print(learner)
#> <LearnerSurvGBM:surv.gbm>: Gradient Boosting
#> * Model: -
#> * Parameters: distribution=coxph, keep.data=FALSE, n.cores=1
#> * Packages: mlr3, mlr3proba, mlr3extralearners, gbm
#> * Predict Types:  [crank], lp
#> * Feature Types: integer, numeric, factor, ordered
#> * Properties: importance, missings, weights

# Define a Task
task = mlr3::tsk("grace")

# Create train and test set
ids = mlr3::partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
#> gbm::gbm(formula = f, distribution = "coxph", data = task$data(), 
#>     weights = NULL, keep.data = FALSE, n.cores = 1L)
#> A gradient boosted model with coxph loss function.
#> 100 iterations were performed.
#> There were 6 predictors of which 6 had non-zero influence.
print(learner$importance())
#> revascdays        age     revasc      sysbp        los   stchange 
#>  40.049166  20.875592  19.065146   9.113981   8.342374   2.553741 

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)
#> Using 100 trees...

# Score the predictions
predictions$score()
#> surv.cindex 
#>   0.8596297