R/method_ebr.R

In rivolli/utiml: Utilities for Multi-Label Learning

#' Ensemble of Binary Relevance for multi-label Classification
#'
#' Create an Ensemble of Binary Relevance model for multilabel classification.
#'
#' This model is composed by a set of Binary Relevance models. Binary Relevance
#' is a simple and effective transformation method to predict multi-label data.
#'
#' @family Transformation methods
#' @family Ensemble methods
#' @param mdata A mldr dataset used to train the binary models.
#' @param base.algorithm A string with the name of the base algorithm. (Default:
#'  \code{options("utiml.base.algorithm", "SVM")})
#' @param m The number of Binary Relevance models used in the ensemble.
#'    (Default: 10)
#' @param subsample A value between 0.1 and 1 to determine the percentage of
#'    training instances that must be used for each classifier. (Default: 0.75)
#' @param attr.space A value between 0.1 and 1 to determine the percentage of
#'    attributes that must be used for each classifier. (Default: 0.50)
#' @param replacement Boolean value to define if use sampling with replacement
#'    to create the data of the models of the ensemble. (Default: TRUE)
#' @param ... Others arguments passed to the base algorithm for all subproblems.
#' @param cores The number of cores to parallelize the training. Values higher
#'  than 1 require the \pkg{parallel} package. (Default:
#'  \code{options("utiml.cores", 1)})
#' @param seed An optional integer used to set the seed. This is useful when
#'  the method is run in parallel. (Default: \code{options("utiml.seed", NA)})
#' @return An object of class \code{EBRmodel} containing the set of fitted
#'   BR models, including:
#' \describe{
#'   \item{models}{A list of BR models.}
#'   \item{nrow}{The number of instances used in each training dataset.}
#'   \item{ncol}{The number of attributes used in each training dataset.}
#'   \item{rounds}{The number of interactions.}
#' }
#' @references
#'    Read, J., Pfahringer, B., Holmes, G., & Frank, E. (2011). Classifier
#'    chains for multi-label classification. Machine Learning, 85(3), 333-359.
#'
#'    Read, J., Pfahringer, B., Holmes, G., & Frank, E. (2009).
#'    Classifier Chains for Multi-label Classification. Machine Learning and
#'    Knowledge Discovery in Databases, Lecture Notes in Computer Science,
#'    5782, 254-269.
#' @note If you want to reproduce the same classification and obtain the same
#'  result will be necessary set a flag utiml.mc.set.seed to FALSE.
#' @export
#'
#' @examples
#' model <- ebr(toyml, "RANDOM")
#' pred <- predict(model, toyml)
#'
#' \donttest{
#' # Use C5.0 with 90% of instances and only 5 rounds
#' model <- ebr(toyml, 'C5.0', m = 5, subsample = 0.9)
#'
#' # Use 75% of attributes
#' model <- ebr(toyml, attr.space = 0.75)
#'
#' # Running in 2 cores and define a specific seed
#' model1 <- ebr(toyml, cores=2, seed = 312)
#' }
ebr <- function(mdata,
                base.algorithm = getOption("utiml.base.algorithm", "SVM"),
                m = 10, subsample = 0.75, attr.space = 0.5, replacement = TRUE,
                ..., cores = getOption("utiml.cores", 1),
                seed = getOption("utiml.seed", NA)) {
  # Validations
  if (!is(mdata, "mldr")) {
    stop("First argument must be an mldr object")
  }

  if (m < 2) {
    stop("The number of iterations (m) must be greater than 1")
  }

  if (subsample < 0.1 || subsample > 1) {
    stop("The subset of training instances must be between 0.1 and 1 inclusive")
  }

  if (attr.space <= 0.1 || attr.space > 1) {
    stop(paste("The attribbute space of training instances must be between ",
          "0.1 and 1 inclusive"))
  }

  if (cores < 1) {
    stop("Cores must be a positive value")
  }

  # EBR Model class
  ebrmodel <- list(rounds = m, call = match.call())
  ebrmodel$nrow <- ceiling(mdata$measures$num.instances * subsample)
  ebrmodel$ncol <- ceiling(length(mdata$attributesIndexes) * attr.space)
  ebrmodel$cardinality <- mdata$measures$cardinality

  if (!anyNA(seed)) {
    set.seed(seed)
  }
  idx <- lapply(seq(m), function(iteration) {
    list(
      rows = sample(mdata$measures$num.instances, ebrmodel$nrow, replacement),
      cols = sample(mdata$attributesIndexes, ebrmodel$ncol)
    )
  })

  ebrmodel$models <- lapply(seq(m), function(iteration) {
    ndata <- create_subset(mdata, idx[[iteration]]$rows, idx[[iteration]]$cols)

    brmodel <- br(ndata, base.algorithm, ..., cores = cores, seed = seed)
    brmodel$attrs <- colnames(ndata$dataset[, ndata$attributesIndexes])
    rm(ndata)

    brmodel
  })

  class(ebrmodel) <- "EBRmodel"
  ebrmodel
}

#' Predict Method for Ensemble of Binary Relevance
#'
#' This method predicts values based upon a model trained by \code{\link{ebr}}.
#'
#' @param object Object of class '\code{EBRmodel}'.
#' @param newdata An object containing the new input data. This must be a
#'  matrix, data.frame or a mldr object.
#' @param vote.schema Define the way that ensemble must compute the predictions.
#'  The default valid options are: c("avg", "maj", "max", "min"). If \code{NULL}
#'  then all predictions are returned. (Default: \code{'maj'})
#' @param probability Logical indicating whether class probabilities should be
#'  returned. (Default: \code{getOption("utiml.use.probs", TRUE)})
#' @param ... Others arguments passed to the base algorithm prediction for all
#'   subproblems.
#' @param cores The number of cores to parallelize the training. Values higher
#'  than 1 require the \pkg{parallel} package. (Default:
#'  \code{options("utiml.cores", 1)})
#' @param seed An optional integer used to set the seed. This is useful when
#'  the method is run in parallel. (Default: \code{options("utiml.seed", NA)})
#' @return An object of type mlresult, based on the parameter probability.
#' @seealso \code{\link[=ebr]{Ensemble of Binary Relevance (EBR)}} \code{
#'    \link[=compute_multilabel_predictions]{Compute Multi-label Predictions}}
#' @export
#'
#' @examples
#' \donttest{
#' # Predict SVM scores
#' model <- ebr(toyml)
#' pred <- predict(model, toyml)
#'
#' # Predict SVM bipartitions running in 2 cores
#' pred <- predict(model, toyml, prob = FALSE, cores = 2)
#'
#' # Return the classes with the highest score
#' pred <- predict(model, toyml, vote = 'max')
#' }
predict.EBRmodel <- function(object, newdata, vote.schema = "maj",
                             probability = getOption("utiml.use.probs", TRUE),
                             ..., cores = getOption("utiml.cores", 1),
                             seed = getOption("utiml.seed", NA)) {
  # Validations
  if (!is(object, "EBRmodel")) {
    stop("First argument must be an EBRmodel object")
  }

  if (cores < 1) {
    stop("Cores must be a positive value")
  }

  utiml_ensemble_check_voteschema(vote.schema)

  newdata <- utiml_newdata(newdata)
  allpreds <- lapply(seq(object$models), function(imodel) {
    brmodel <- object$models[[imodel]]
    predict.BRmodel(brmodel, newdata[, brmodel$attrs], ...,
                    cores = cores, seed = seed)
  })

  prediction <- utiml_predict_ensemble(allpreds, vote.schema, probability)
  if (!is.null(vote.schema)) {
    prediction <- lcard_threshold(prediction, object$cardinality, probability)
  }
  prediction
}

#' Print EBR model
#' @param x The ebr model
#' @param ... ignored
#'
#' @return No return value, called for print model's detail
#'
#' @export
print.EBRmodel <- function(x, ...) {
  cat("Ensemble of Binary Relevance Model\n\nCall:\n")
  print(x$call)
  cat("\nDetails:")
  cat("\n ", x$rounds, "Iterations")
  cat("\n ", x$nrow, "Instances")
  cat("\n ", x$ncol, "Attributes\n")
}