mlr3superlearner {mlr3superlearner}R Documentation

Super Learner Algorithm

Description

Implementation of the Super Learner algorithm using the 'mlr3' framework. By default, returning the discrete Super Learner. If using the ensemble Super Learner, The LASSO with an alpha value of 0 and a restriction on the lower limit of the coefficients is used as the meta-learner.

Usage

mlr3superlearner(
  data,
  target,
  library,
  outcome_type = c("binomial", "continuous"),
  folds = NULL,
  discrete = TRUE,
  newdata = NULL,
  group = NULL,
  info = FALSE
)

Arguments

data

[data.frame]
A data.frame containing predictors and target variable.

target

[character(1)]
The name of the target variable in data.

library

[character] or [list]
A vector or list of algorithms to be used for prediction.

outcome_type

[character(1)]
The outcome variable type. Options are "binomial" and "continuous".

folds

[numeric(1)]
The number of cross-validation folds, or if NULL will be dynamically determined.

discrete

[logical(1)]
Return the discrete Super Learner, or the ensemble Super Learner?

newdata

[list]
A list of data.frames to generate predictions from.

group

[character(1)]
Name of a grouping variable in data. Assumed to be discrete; observations in the same group are treated like a "block" of observations kept together during sample splitting.

info

[logical(1)]
Print learner fitting information to the console.

Value

A list of class mlr3superlearner.

Examples

if (requireNamespace("ranger", quietly = TRUE)) {
  n <- 1e3
  W <- matrix(rnorm(n*3), ncol = 3)
  A <- rbinom(n, 1, 1 / (1 + exp(-(.2*W[,1] - .1*W[,2] + .4*W[,3]))))
  Y <- rbinom(n,1, plogis(A + 0.2*W[,1] + 0.1*W[,2] + 0.2*W[,3]^2 ))
  tmp <- data.frame(W, A, Y)
  mlr3superlearner(tmp, "Y", c("glm", "ranger"), "binomial")
}

[Package mlr3superlearner version 0.1.2 Index]