R/wfpi.R

In jaroslav-kuchar/fpmoutliers: Frequent Pattern Mining Outliers

#' WFPI - Weighted Frequent Pattern Isolation algorithm
#'
#' Algorithm proposed by:
#' J. Kuchar, V. Svatek: Spotlighting Anomalies using Frequent Patterns, Proceedings of the KDD 2017 Workshop on Anomaly Detection in Finance, Halifax, Nova Scotia, Canada, PMLR, 2017.
#'
#' @param data \code{data.frame} or \code{transactions} from \code{arules} with input data
#' @param minSupport minimum support for FPM
#' @param mlen maximum length of frequent itemsets
#' @param preferredColumn column name that is preferred
#' @param preference numeric value that multiplies the score
#' @param noCores number of cores for parallel computation
#' @return model output (list) with all results including outlier scores
#' @import arules foreach doParallel parallel
#' @export
#' @examples
#' library("fpmoutliers")
#' dataFrame <- read.csv(
#'      system.file("extdata", "fp-outlier-customer-data.csv", package = "fpmoutliers"))
#' model <- WFPI(dataFrame, minSupport = 0.001, preferredColumn="Car", preference=10)
WFPI <- function(data, minSupport=0.3, mlen=0, preferredColumn="", preference=1.0, noCores=1){
  registerDoParallel(noCores)

  if(is(data,"data.frame")){
    data <- sapply(data,as.factor)
    data <- data.frame(data, check.names=F)
    txns <- as(data, "transactions")
  } else {
    txns <- data
  }
  if(mlen<=0){
    variables <- unname(sapply(txns@itemInfo$labels,function(x) strsplit(x,"=")[[1]][1]))
    mlen <- length(unique(variables))
  }
  fitemsets <- apriori(txns, parameter = list(support=minSupport, maxlen=mlen, target="frequent itemsets"))

  fiList <- LIST(items(fitemsets))
  qualities <- fitemsets@quality[,"support"]

  scores <- c()
  tx <- NULL
  scores <- foreach(tx = as(txns,"list"), .combine = list, .multicombine = TRUE)  %dopar%  {
    transaction = unlist(tx,"list")

    coverage <- c()
    support <- c()
    for(item in seq(1,length(fiList))){
      itemset <- fiList[[item]]
      if(all(itemset %in% transaction)){
        if(preferredColumn!="" && any(grepl(paste(preferredColumn,"=",sep=""), itemset))) {
          support <- c(support, preference*1/(qualities[item]*length(itemset)))
        } else {
          support <- c(support, 1/(qualities[item]*length(itemset)))
        }
        coverage <- unique(c(coverage, itemset))
      }
    }
    # penalization for incomplete coverage
    if(length(coverage)<length(transaction)){
      support <- c(support, rep(length(txns),(length(transaction) - length(coverage)) ))
    }

    if(length(support)>0){
      mean(support)
    } else {
      length(txns)
    }
  }
  scores <- unlist(scores)
  stopImplicitCluster()

  output <- list()
  output$minSupport <- minSupport
  output$maxlen <- mlen
  output$model <- fitemsets
  output$scores <- scores
  output
}