R/gb_classifier.R

In autoMrP: Improving MrP with Ensemble Learning

#' GB classifier
#'
#' \code{gb_classifier} applies gradient boosting classification to a data set.
#'
#' @inheritParams auto_MrP
#' @param form Model formula. A two-sided linear formula describing
#'   the model to be fit, with the outcome on the LHS and the covariates
#'   separated by + operators on the RHS.
#' @param distribution Model distribution. A character string specifying the
#'   name of the distribution to be used.
#' @param data.train Training data. A data.frame containing the training data
#'   used to train the model.
#' @param n.trees Total number of trees. An integer-valued scalar specifying
#'   the total number of trees to be fit.
#' @param interaction.depth Interaction depth. An integer-valued scalar
#'   specifying the maximum depth of each tree.
#' @param n.minobsinnode Minimum number of observations in terminal nodes. An
#'   integer-valued scalar specifying the minimum number of observations in the
#'   terminal nodes of the trees.
#' @param shrinkage Learning rate. A numeric scalar specifying the shrinkage or
#'   learning rate applied to each tree in the expansion.
#' @param verbose Verbose output. A logical vector indicating whether or not
#'   verbose output should be printed.
#' @return A gradient tree boosting model. A \code{\link[gbm]{gbm}} object.

gb_classifier <- function(
  y, form, distribution, data.train,
  n.trees, interaction.depth,
  n.minobsinnode, shrinkage,
  verbose = c(TRUE, FALSE)
) {

  # Determine type of dependent variable
  if (
    data.train %>%
      dplyr::pull(!!y) %>%
      unique() %>%
      length() > 2
  ) {
    # set model family to gaussian
    distribution <- "gaussian"
  }

  # Train model on training data with number of total trees, interaction depth,
  # and learning rate as tuning parameters
  if (isTRUE(verbose == TRUE)) {
    out <- gbm::gbm(
      formula = form,
      distribution = distribution,
      data = data.train,
      n.trees = n.trees,
      interaction.depth = interaction.depth,
      n.minobsinnode = n.minobsinnode,
      shrinkage = shrinkage,
      train.fraction = 1,
      n.cores = 1
    )
  } else {
    out <- suppressMessages(suppressWarnings(
      gbm::gbm(
        formula = form,
        distribution = distribution,
        data = data.train, n.trees = n.trees,
        interaction.depth = interaction.depth,
        n.minobsinnode = n.minobsinnode,
        shrinkage = shrinkage,
        train.fraction = 1,
        n.cores = 1
      )
    ))
  }

  # Function output
  return(out)
}

#' GB classifier update
#'
#' \code{gb_classifier_update()} grows additional trees in gradient tree
#' boosting ensemble.
#'
#' @param object Gradient tree boosting output. A gbm object.
#' @param n.new.trees Number of additional trees to grow. A numeric scalar.
#' @param verbose Verbose output. A logical vector indicating whether or not
#'   verbose output should be printed.
#' @return An updated gradient tree boosting model.
#'   A \code{\link[gbm]{gbm.more}} object.

gb_classifier_update <- function(
  object, n.new.trees, verbose = c(TRUE, FALSE)
) {

  # Train model on training data with number of total trees, interaction depth,
  # and learning rate as tuning parameters
  if (isTRUE(verbose == TRUE)) {
    out <- gbm::gbm.more(
      object = object,
      n.new.trees = n.new.trees
    )
  } else {
    out <- suppressMessages(suppressWarnings(
      gbm::gbm.more(
        object = object,
        n.new.trees = n.new.trees
      )
    ))
  }

  # Function output
  return(out)
}