R/mlr.learners.classif.rpart.R

In mlr-org/mlrng: Machine Learning in R - Next Generation

#' @include mlr.learners.R
#' @include LearnerClassif.R
mlr.learners$add(LearnerClassif$new(
  name = "rpart",
  package = "rpart",
  par.set = ParamSetFlat$new(params = list(
    ParamInt$new(id = "minsplit", default = 20L, lower = 1L),
    ParamReal$new(id = "cp", default = 0.01, lower = 0, upper = 1),
    ParamInt$new(id = "maxcompete", default = 4L, lower = 0L),
    ParamInt$new(id = "maxsurrogate", default = 5L, lower = 0L),
    #FIXME: disabled due to phng problem, but rpart needs to be removed anyhow
    # ParamCategorical$new(id = "usesurrogate", default = 2L, values = 0:2),
    # ParamCategorical$new(id = "surrogatestyle", default = 0L, values = 0:1),
    # we use 30 as upper limit, see docs of rpart.control
    ParamInt$new(id = "maxdepth", default = 30L, lower = 1L, upper = 30L),
    ParamInt$new(id = "xval", default = 10L, lower = 0L)
  )),
  par.vals = list(),
  properties = c("twoclass", "multiclass", "missings", "feat.numeric", "feat.factor", "feat.ordered", "prob", "weights", "featimp", "formula"),

  train = function(task, subset, ...) {
    data = getTaskData(task, subset = subset, type = "train", props = self$properties)
    rpart::rpart(task$formula, data, ...)
  },

  predict = function(model, newdata, ...) {
    pt = self$predict.type
    if (pt == "response")
      as.character(predict(model$rmodel, newdata = newdata, type = "class", ...))
    else
        predict(model$rmodel, newdata = newdata, type = "prob", ...)
  },
  # FIXME: can be removed, was just for testing
  model.extractors = list(residuals = function(model, task, subset, type = "usual", ...) {
    type = assertSubset(type, choices = c("usual", "pearson", "deviance"))
    rpart:::residuals.rpart(model, type = type, ...)
  })
))