R/multistate_test.R

In KNewhart/ADPCA: ADPCA for Process Monitoring of WWTP

#' Test based on multistate data
#'
#' This function tests data for multistate AD-PCA
#' @param data compiled and cleaned data
#' @param trainingSpec_ls ADPCA training specifications
#' @param testingDay day to be tested
#' @param faultsToTriggerAlarm number of faults to trigger an alarm (defaults to 5)
#' @return alarms.xts
#' @export


multistate_test <- function(data, trainingSpec_ls, testingDay, faultsToTriggerAlarm = 5) {

  # Creates list of the size of the number of states from stateGenerator
  alarmData_ls <- list()
  if (is.list(trainingSpec_ls)) {
    l <- length(trainingSpec_ls)
  } else {
    l <- 1
  }


  for (i in 1:l) {
  # alarmData_ls <- lapply(list(seq(1,l)), function(i) {
    if(length(trainingSpec_ls[[i]]) == 0) {
      alarmData_ls <- c(alarmData_ls, list(NULL))
      next
    }
    alarmDataHolder <- ADPCA::testNewObs(data = data[[i]],
                      trainingSpecs = trainingSpec_ls[[i]],
                      testingDay = testingDay,
                      faultsToTriggerAlarm = faultsToTriggerAlarm)
    alarmData_ls <- c(alarmData_ls, list(alarmDataHolder))
  }

  # Compile all states
  alarmData <- data.frame()
  for (i in 1:length(alarmData_ls)) {
    if (i == 1) {
      alarmData <- as.data.frame(alarmData_ls[[i]])
    } else {
      alarmData <- fastmerge(alarmData, as.data.frame(alarmData_ls[[i]]))
    }
  }

  timestamps <- as.POSIXct(rownames(alarmData), format = "%Y-%m-%d %H:%M:%S")
  alarms.xts <- as.xts(alarmData, order.by = timestamps)

  # Return xts with test data
  return(alarms.xts)

}