R/predict.DECORATE.R

In DecorateR: Fit and Deploy DECORATE Trees

#' Predict method for DECORATE objects
#'
#' Prediction of new data using DECORATE
#'
#' @param object an object of the class \code{DECORATE}, as created by the function \code{DECORATE}.
#' @param newdata a data frame containing the same predictors as in the training phase.
#' @param type character specifying whether to return the probabilites ('prob') or class ('class'). Default: prob.
#' @param all Return the predictions per tree instead of the average (default = FALSE).
#' @param ... Not used currently.
#'
#' @return  vector containing the response probabilities.
#'
#'
#' @examples
#'
#' data(iris)
#' y <- as.factor(ifelse(iris$Species[1:100]=="setosa",0,1))
#' x <- iris[1:100,-5]
#' dec <- DECORATE(x = x, y = y)
#' predict(object=dec,newdata=x)
#'
#'


predict.DECORATE <- function (object, newdata, type = 'prob',
                              all =FALSE, ...) {
  final <- newdata

  if (!is.data.frame(final)) stop("newdata must be a data frame")

  predicted <- matrix(NA,nrow=nrow(final),ncol=length(object))

  if(type == 'prob') {
    for (i in 1:length(object)){
      predicted[,i] <-predict(object[[i]], final, 'prob')[,2]
    }

    if(all) {
      return(predicted)
    } else {
      finalpredicted <- rowMeans(predicted)
      return(finalpredicted)
    }
  }

  if(type == 'class') {

    Mode <- function(x) {
      ux <- unique(x)
      ux[which.max(tabulate(match(x, ux)))]
    }

    for (i in 1:length(object)){
      predicted[,i] <- as.numeric(as.character(predict(object[[i]], final, 'class')))
    }

    if(all) {
      return(predicted)
    } else {
      finalpredicted <- as.factor(apply(predicted, 1, Mode))
      return(finalpredicted)
    }
  }

}