R/diagnose.R

In TomaszKoska/RUtils:

diagnose <- function(fo,minObs=10){
  #funkcja weźmie fo i sprawdzi:
  # w każdym data secie
  #wszystko to co diagnoseHelper robi
  #i inne rzeczy też...

  if(class(fo) != "ForecastingObject"){
    warning("This function should get ForecastingObject as parameter. Please use buildForecastingObject.")
    return(NULL)
  }



  if(fo$yName %in% names(fo$train)){
    print(paste(c("y variable ", fo$yName , " exists in train dataset!"),collapse=""))
  }
  if(fo$yName %in% names(fo$trainFull)){
    print(paste(c("y variable ", fo$yName, " exists in trainFull dataset!"),collapse=""))
  }
  if(fo$yName %in% names(fo$forecast)){
      print(paste(c("y variable ", fo$yName, " exists in forecast dataset!"),collapse=""))
    }

  tDiffs<-setdiff(names(fo$train),names(fo$forecast))
  fDiffs<- setdiff(names(fo$forecast),names(fo$train))
  trainDiffs <- paste(tDiffs,collapse = ",")
  forecastDiff <- paste(fDiffs,collapse = ",")

  if(length(tDiffs) > 0){
    print(paste(c("Variables missing in train that are in forecast: ", trainDiffs),collapse=""))
  }
  if(length(fDiffs) > 0){
    print(paste(c("Variables missing in train that are in forecast: ", forecastDiff),collapse=""))
  }




  print("Train dataset")
  diagnoseHelper(fo$train,fo$yName,minObs)
  print("")
  print("")
  print("")
  print("Full train dataset")
  diagnoseHelper(fo$trainFull,fo$yName,minObs)
  print("")
  print("")
  print("")
  print("Forecast dataset")
  diagnoseHelper(fo$forecast,fo$yName,minObs)
}

diagnoseHelper <- function(df,yName,minObs){
  require(caret)
  # 1. ile jest wierszy?
  # 2. ile jest kolumn?
  # 3. w ilu wierszach brakuje obserwacji?
  # 4. czy istnieją zmienne które nie mają wariancji (i które to)?
  # 5. czy istniją zmienne, w których level jest nieużywany? <- TODO!!!!
  df <- df[,setdiff(names(df),yName)]
  print(paste(c("There is ", nrow(df), " rows!"),collapse=""))
  print(paste(c("There is ", ncol(df), " columns!"),collapse=""))
  x<- apply(df,2,pMiss)
  x<-x[x>0]
  if(length(x)>0){
    for(a in 1:length(x)){
      print(paste(c("There is ", x[a], " missing values in ", names(x)[a] ," variable!"),collapse=""))
    }
  }else{
    print("No missing values!")
  }

  x <- nearZeroVar(df, saveMetrics= TRUE)
  x <- x[x$zeroVar==TRUE,]

  for(a in row.names(x)){
    print(paste(c("Variable ", a ," has no variance!"),collapse=""))
  }

  x<- sapply(df, class)
  x<-names(x[x=="factor"])
  a<-apply(df[,x],2,table)

  i<-0
  for(v in a){
    i<-i+1
    # print(v)
    y <-v[v<=minObs]
    if(length(y)>0){
      # print(x)
      print(paste(c("Variable ", x[i] ," has some rare levels!"),collapse=""))

    }
  }
}