This function is used to randomly split a Netica casefile into K splits for K-fold cross validation. K is specified by the user. Seed is also set by the user for reproducibility. Path1 and Path2 are character file paths to the directory of the write.paths for the training and holdout case files, respectively. File path arguments should end with …

KfoldCASEFILE.Rd

## FUNCTION - KfoldCASEFILE
# splits an R object of a case file into 5 Folds (When K is set to 5)
# Outputs the object into five pairs (When K = 5) of training and test case files in .cas format
# The input case file object must contain a column called "IDnum"
KfoldCASEFILE<-function(CaseFile, seed,k, path1, path2){
  # Renumber the IDnum column by order IF the case file is missing a number in a sequence
  CaseFile<-IDnumOrdered(CaseFile)
  #1. Setting the folds
  #2. Join the folds to the casefile object by dplyr::innerjoin (by cols are index in folds and idnum in casefile object)
  set.seed(seed)
  Folds<-cvTools::cvFolds(nrow(transect.lookup.table), K = k, type = "random")
  Folds<-data.frame(Folds$subsets, Folds$which) %>%
    dplyr::rename(IDnum =Folds.subsets, Fold = Folds.which) %>%
    dplyr::inner_join(.,CaseFile)

  #Set up training and holdout vectors
  splits<-data_frame(c(1:4, 5),
                     c(1:3,5, 4),
                     c(1:2, 4:5, 3),
                     c(1,3:5,2),
                     c(2:5, 1))
  colnames(splits)<-paste0("split",c(1:5))
  splits$which<-c(rep("training",4),"holdout")
  # set output file paths
  training.casefile.out<- path1
  holdout.casefile.out<-path2
  #Create the case files, 2 for every column in splits except for "which"
  for(i in seq_len(k)){
    vec<-select(splits, i, which) #take the relevent column vectors (split1:5), and the which col
    split.name<-names(splits)[i]
    names(vec)[1]<-"Fold" #rename the first col to allow joining
    training.holdout<-dplyr::inner_join(Folds, vec) # join vec to Folds by Fold (automatically done)
    #Subset folds twice and store inside objects named after sets 1 and 2 respectively
    training<-filter(training.holdout, which == "training") %>%
      select(-Fold,-which)
    holdout<-filter(training.holdout, which == "holdout") %>%
      select(-Fold,-which)
    #Write both of the casefiles
    training.casefile.path<-paste(training.casefile.out,split.name, "training",".cas", sep="" )
    holdout.casefile.path<-paste(holdout.casefile.out,split.name, "holdout",".cas", sep="" )
    CaseFileDelimiter("\t")
    CaseFileMissingCode("*")
    write.table(training, file = training.casefile.path, append = FALSE, quote = FALSE, sep = CaseFileDelimiter(),
                eol = "\n", na = CaseFileMissingCode(), dec = ".", row.names = FALSE,
                col.names = TRUE, qmethod = c("escape", "double"),
                fileEncoding = "")
    write.table(holdout, file = holdout.casefile.path, append = FALSE, quote = FALSE, sep = CaseFileDelimiter(),
                eol = "\n", na = CaseFileMissingCode(), dec = ".", row.names = FALSE,
                col.names = TRUE, qmethod = c("escape", "double"),
                fileEncoding = "")
  }
}