R/train_basemodel_core.R

In stacking: Building Predictive Models with Stacking

train_basemodel_core <- function(repeat.parLapply, division, l, core, x, y, exclude){

  x.train <- x[-exclude, ]
  y.train <- y[-exclude]

  cl <- makeCluster(core, type="SOCK")
  clusterExport(cl, c("x.train",
                      "y.train",
                      "train"), envir = environment())

  train_result <- NULL
  for(rp in 1:repeat.parLapply){

    train_result <- c(train_result,
                      parLapply(cl, l[division[, rp]],
                                function(m){
                                  result <- train(x.train, y.train, method = m$method, tuneGrid = m$hyp)
                                  result$trainingData <- NULL
                                  result
                                })
    )
  }
  stopCluster(cl)

  train_result
}