Boosted Generalized Linear Survival Learner

Fits a generalized linear survival model using a boosting algorithm. Calls mboost::glmboost() from mboost.

Dictionary

This Learner can be instantiated via lrn():

lrn("surv.glmboost")

Meta Information

• Task type: “surv”

• Predict Types: “crank”, “distr”, “lp”

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

• Required Packages: mlr3, mlr3proba, mlr3extralearners, mboost, pracma

Parameters

Id	Type	Default	Levels	Range
offset	numeric	-		\((-\infty, \infty)\)
family	character	coxph	coxph, weibull, loglog, lognormal, gehan, cindex, custom	-
custom.family	untyped	-		-
nuirange	untyped	c(0, 100)		-
center	logical	TRUE	TRUE, FALSE	-
mstop	integer	100		\([0, \infty)\)
nu	numeric	0.1		\([0, 1]\)
risk	character	inbag	inbag, oobag, none	-
oobweights	untyped	NULL		-
stopintern	logical	FALSE	TRUE, FALSE	-
trace	logical	FALSE	TRUE, FALSE	-
sigma	numeric	0.1		\([0, 1]\)
ipcw	untyped	1		-
na.action	untyped	stats::na.omit		-
contrasts.arg	untyped	-		-

Prediction types

This learner returns two to three prediction types:

1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using mboost::predict.glmboost(). If the family parameter is not "coxph", -lp is returned, since non-coxph families represent AFT-style distributions where lower lp values indicate higher risk.

2. crank: same as lp.

3. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using mboost::survFit(). This prediction type is present only when the family distribution parameter is equal to "coxph" (default). By default the Breslow estimator is used for computing the baseline hazard.

References

Bühlmann, Peter, Yu, Bin (2003). “Boosting with the L 2 loss: regression and classification.” Journal of the American Statistical Association, 98(462), 324–339.

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 -> LearnerSurvGLMBoost

Methods

Public methods

• LearnerSurvGLMBoost$new()

• LearnerSurvGLMBoost$importance()

• LearnerSurvGLMBoost$selected_features()

• LearnerSurvGLMBoost$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

LearnerSurvGLMBoost$new()

---

Method importance()

Importance scores are extracted with the function mboost::varimp() and represent a feature's individual contribution to the risk reduction per boosting step of the fitted model. The higher the risk reduction, the larger the feature importance.

Note: Importance is supported only for datasets with numeric features, as the presence of factors with multiple levels makes it difficult to get the original feature names.

Usage

LearnerSurvGLMBoost$importance()

Returns

Named numeric().

---

Method selected_features()

Selected features are extracted with the function mboost::coef.glmboost() which by default returns features with non-zero coefficients and for the final number of boosting iterations.

Note: Selected features can be retrieved only for datasets with numeric features, as the presence of factors with multiple levels makes it difficult to get the original feature names.

Usage

LearnerSurvGLMBoost$selected_features()

Returns

character().

---

Method clone()

The objects of this class are cloneable with this method.

Usage

LearnerSurvGLMBoost$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

# Define the Learner
learner = mlr3::lrn("surv.glmboost")
print(learner)
#> <LearnerSurvGLMBoost:surv.glmboost>: Boosted Generalized Linear Model
#> * Model: -
#> * Parameters: family=coxph
#> * Packages: mlr3, mlr3proba, mlr3extralearners, mboost, pracma
#> * Predict Types:  [crank], distr, lp
#> * Feature Types: logical, integer, numeric, factor
#> * Properties: importance, selected_features, 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)
#> Warning: model with centered covariates does not contain intercept

print(learner$model)
#> 
#> 	 Generalized Linear Models Fitted via Gradient Boosting
#> 
#> Call:
#> glmboost.matrix(x = as.matrix(task$data(cols = task$feature_names)),     y = task$truth(), family = family)
#> 
#> 
#> 	 Cox Partial Likelihood 
#> 
#> Loss function:  
#> 
#> Number of boosting iterations: mstop = 100 
#> Step size:  0.1 
#> Offset:  0 
#> 
#> Coefficients: 
#>          age       revasc   revascdays 
#>  0.017849006 -0.944969504 -0.007633255 
#> attr(,"offset")
#> [1] 0
#> 
print(learner$importance())
#> revascdays     revasc        age        los   stchange      sysbp 
#> 0.06936821 0.05656239 0.03622781 0.00000000 0.00000000 0.00000000 

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
#> surv.cindex 
#>   0.8048688