#clear everything
rm(list = setdiff(ls(), lsf.str()))
library(caret)

BasePath = "D:/initialexperiments/initialexperiments/"
PathOfDataSet = paste0(BasePath, "6-Scaled Dataset/")
dataSetName = datasetRepository = list.files(path = PathOfDataSet)

totalbreaks = 10
size = 50

serial <- 0
SrNo <- integer(0)
datasets <- character(0)
featureQuantity <- integer(0)
featureList <- character(0)
instances <- integer(0)
faultyInstances <- integer(0)
totalTPs <- integer(0)
totalTNs <- integer(0)
totalFNs <- integer(0)
totalFPs <- integer(0)
selectedModels <- list()
hiddenUnits <- str(0)
foldTPs <- str(0)
foldTNs <- str(0)
foldFNs <- str(0)
foldFPs <- str(0)

library(nnet) #for confusion metrix
modelQuantity <- 0

for (names in 1:NROW(dataSetName)) {
  names <- 1
  CompletePathAndFile = paste0(PathOfDataSet, dataSetName[names])
  mydata = read.csv(CompletePathAndFile, header = TRUE)
  mydata$BUG <- as.numeric(mydata$BUG)
  currentDataSet <- mydata
  
  #Randomly shuffle the data 10 times
  for (k in 1:10) {
    currentDataSet <-
      currentDataSet[sample(nrow(currentDataSet)), ]
  }
  folds <- createFolds(currentDataSet$BUG, k = totalbreaks)
  TPs <- integer(0)
  FPs <- integer(0)
  FNs <- integer(0)
  TNs <- integer(0)
  hiddenUnit <- integer(0)
  for (l in 1:totalbreaks) {
    l <- 1
    modelQuantity <- modelQuantity + 1
    testIndexes <- unlist(folds[l])
    testData <- currentDataSet[testIndexes, ]
    trainData <- currentDataSet[-testIndexes, ]
    set.seed(200)
    ######################## model building ###################
    n <- colnames(trainData)
    
    f <-      as.formula(paste("BUG ~", paste(n[!n %in% "BUG"], collapse = " + ")))
    cat("model building...")
    values <- integer(0)
    myModels <- list()
    modelLength <- 0
    for (i in seq(1, size, 1)) {
      i <- 10
      cat(
        "in process---------------------------------------------",
        i,
        "--",
        names,
        "Folds",
        l,
        "\n"
      )
      modelLength <- modelLength + 1
      myModel <-
        nnet(f,
             data = trainData,
             linear.output = F,
             size = i)
      myModels[[modelLength]] <- myModel
      values <- append(values, myModel$value)
    }
    myModel <-  myModels[[match(min(values), values)]]
    selectedModels[[modelQuantity]] <- myModel
    cat("model built...")
    test.predictions <- predict(myModel, testData)
    ########################model building ###################
    results <- ifelse(test.predictions > 0.5, 1, 0)
    original_values <- testData[, ncol(testData)]
    
    length(which(results == original_values))
    
    # ordering the level of the prediction model, so that cf would work
    TP <- length(which(results == original_values & results == 1))
    FN <- length(which(original_values == 1 & results == 0))
    FP <- length(which(original_values == 0 & results == 1))
    TN <- length(which(results == original_values & results == 0))
    TPs <- append(TPs,  TP)
    TNs <- append(TNs,  TN)
    FPs <- append(FPs,  FP)
    FNs <- append(FNs,  FN)
    hiddenUnit <- append(hiddenUnit, myModel$nunits)
  }
  currentFeature <-
    paste(sort(colnames(currentDataSet)[-length(colnames(currentDataSet))]), collapse = ",")
  serial <- serial + 1
  
  SrNo <- append(SrNo, serial)
  datasets <- append(datasets, dataSetName[names])
  featureQuantity <-
    append(featureQuantity, length(colnames(currentDataSet[-length(colnames(currentDataSet))])))
  featureList <-
    append(featureList, paste(sort(colnames(currentDataSet)[-length(colnames(currentDataSet))]), collapse = ","))
  instances <- append(instances, nrow(currentDataSet))
  faultyInstances <-
    append(faultyInstances, prop.table(table(currentDataSet$BUG))[2])
  totalTPs <- append(totalTPs,   sum(TPs))
  totalTNs <- append(totalTNs,   sum(TNs))
  totalFNs <- append(totalFNs,   sum(FNs))
  totalFPs <- append(totalFPs,   sum(FPs))
  

  foldTPs <- append(foldTPs,  paste(as.character(TPs), collapse = ';'))
  foldTNs <- append(foldTNs,  paste(as.character(TNs), collapse = ';'))
  foldFNs <- append(foldFNs,  paste(as.character(FNs), collapse = ';'))
  foldFPs <- append(foldFPs,  paste(as.character(FPs), collapse = ';'))
  
  hiddenUnits <-
    append(hiddenUnits, paste(as.character(hiddenUnit), collapse = ';'))
}

df <- data.frame(
  SrNo,
  datasets,
  featureQuantity,
  featureList,
  instances,
  faultyInstances,
  totalTPs,
  totalTNs,
  totalFNs,
  totalFPs,
  foldTPs,
  foldTNs,
  foldFNs,
  foldFPs,
  hiddenUnits
  
)
write.csv(df,
          row.names = FALSE,
          file =  paste0(BasePath, "Results/",
                         "NNResult6.csv"))

##################################### Results synthesis