R/predict.R

In bmihaljevic/bnclassify: Learning Discrete Bayesian Network Classifiers from Data

#'Predicts class labels or class posterior probability distributions.
#'
#'@details Ties are resolved randomly. Inference is much slower if
#'\code{newdata} contains \code{NA}s.
#'
#'@export
#'
#'@param object A \code{\link{bnc_bn}} object.
#'@param newdata A data frame containing observations whose class has to be 
#'  predicted.
#'@param prob A logical. Whether class posterior probability should be returned.
#'@param ... Ignored.
#'@return If \code{prob=FALSE}, then returns a length-\eqn{N} factor with the 
#'  same levels as the class variable in \code{x}, where \eqn{N} is the number 
#'  of rows in \code{newdata}. Each element is the most likely 
#'  class for the corresponding row in \code{newdata}. If \code{prob=TRUE}, 
#'  returns a \eqn{N} by \eqn{C} numeric matrix, where \eqn{C} is the number of 
#'  classes; each row corresponds to the class posterior of the instance.
#'  
#'@examples 
#'data(car)
#'nb <- bnc('nb', 'class', car, smooth = 1)
#'p <- predict(nb, car)
#'head(p)
#'p <- predict(nb, car, prob = TRUE)
#'head(p)
predict.bnc_fit <- function(object, newdata, prob = FALSE, ...) {      
  pred <- compute_cp(x = object, dataset = newdata, ...)  
  if (!prob) {
    pred <- map(pred)
  }
  pred
}
#' Assigns instances to the most likely class.
#' 
#' Ties are resolved randomly.
#' 
#' @param pred A numeric matrix. Each row corresponds to class posterior 
#'   probabilities for an instance.
#' @return a factor with the same levels as the class variable.
#' @keywords internal
map <- function(pred) {
  max_ind <- max.col(m = pred, ties.method = "random") 
  classes <- colnames(pred)
  stopifnot(is_non_empty_complete(classes))
  predicted <- classes[max_ind]
  #   Return a factor with the levels of the class variable
  factor(predicted, levels = classes)
}