#' @export
makeRLearner.regr.ranger = function() {
makeRLearnerRegr(
cl = "regr.ranger",
package = "ranger",
par.set = makeParamSet(
makeIntegerLearnerParam(id = "num.trees", lower = 1L, default = 500L),
# FIXME: Add default value when data dependent defaults are implemented: mtry=floor(sqrt(#independent vars))
makeIntegerLearnerParam(id = "mtry", lower = 1L),
makeIntegerLearnerParam(id = "min.node.size", lower = 1L, default = 5L),
makeLogicalLearnerParam(id = "replace", default = TRUE),
makeNumericLearnerParam(id = "sample.fraction", lower = 0L, upper = 1L),
makeNumericVectorLearnerParam(id = "split.select.weights", lower = 0, upper = 1),
makeUntypedLearnerParam(id = "always.split.variables"),
makeLogicalLearnerParam(id = "respect.unordered.factors", default = FALSE),
makeDiscreteLearnerParam(id = "importance", values = c("none", "impurity", "permutation"), default = "none", tunable = FALSE),
makeLogicalLearnerParam(id = "write.forest", default = TRUE, tunable = FALSE),
makeLogicalLearnerParam(id = "scale.permutation.importance", default = FALSE, requires = quote(importance == "permutation"), tunable = FALSE),
makeIntegerLearnerParam(id = "num.threads", lower = 1L, when = "both", tunable = FALSE),
makeLogicalLearnerParam(id = "save.memory", default = FALSE, tunable = FALSE),
makeLogicalLearnerParam(id = "verbose", default = TRUE, when = "both", tunable = FALSE),
makeIntegerLearnerParam(id = "seed", when = "both", tunable = FALSE),
makeDiscreteLearnerParam(id = "splitrule", values = c("variance", "maxstat"), default = "variance"),
makeNumericLearnerParam(id = "alpha", lower = 0L, upper = 1L, default = 0.5, requires = quote(splitrule == "maxstat")),
makeNumericLearnerParam(id = "minprop", lower = 0L, upper = 1L, default = 0.1, requires = quote(splitrule == "maxstat")),
makeLogicalLearnerParam(id = "keep.inbag", default = FALSE, tunable = FALSE)
),
par.vals = list(num.threads = 1L, verbose = FALSE, respect.unordered.factors = TRUE),
properties = c("numerics", "factors", "ordered", "oobpreds", "featimp"),
name = "Random Forests",
short.name = "ranger",
note = "By default, internal parallelization is switched off (`num.threads = 1`), `verbose` output is disabled, `respect.unordered.factors` is set to `TRUE`. All settings are changeable.",
callees = "ranger"
)
}
#' @export
trainLearner.regr.ranger = function(.learner, .task, .subset, .weights, ...) {
tn = getTaskTargetNames(.task)
ranger::ranger(formula = NULL, dependent.variable = tn, data = getTaskData(.task, .subset), ...)
}
#' @export
predictLearner.regr.ranger = function(.learner, .model, .newdata, ...) {
p = predict(object = .model$learner.model, data = .newdata, ...)
return(p$predictions)
}
#' @export
getOOBPredsLearner.regr.ranger = function(.learner, .model) {
.model$learner.model$predictions
}
#' @export
getFeatureImportanceLearner.regr.ranger = function(.learner, .model, ...) {
getFeatureImportanceLearner.classif.ranger(.learner, .model, ...)
}
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.