Upload New File

LoMSTEDP.R

+#Following function implements the LoMST algorithm and returns the anomaly indices and number of true detection for a preselected k value and a cut-off value
+LoMSTEDP=function(selectedk){
+  nn <- kNN(data, k=selectedk)
+  kn=as.matrix(nn$id)
+  
+  for(i in 1:nrow(data)) 
+  { 
+    nam<- paste("dist", i, sep = "")
+    assign(nam, as.matrix(dist(data[c(i,kn[i,]),])))
+  }
+  
+  sumhist=numeric(nrow(data))
+  compare=numeric(nrow(data))
+  for(i in 1:nrow(data)){
+    d=get(paste("dist", i, sep = ""))
+    mat=as.matrix(dino.mst(d))
+    d[which(mat== 0)] <- 0
+    
+    sumhist[i]=sum(d)}
+  for(i in 1:nrow(data)){
+    compare[i]=sumhist[i]-mean(sumhist[c(kn[i,])])}
+  
+  summary=as.data.frame(compare)
+  summary$obs=c(1:nrow(data))
+  sumhistarranged=as.data.frame(arrange(summary,desc(compare)))
+  #listofoutliers=sumhistarranged$obs[1:N]
+  dat$d=c(1:nrow(data))
+  required=dat[,c(1,ncol(dat))]
+  names(required)=c("timestamp","obs")
+  final=as.data.frame(merge(sumhistarranged,required,by=c("obs")))
+  final$Outlier_Score=(final$compare-min(final$compare))/(max(final$compare)-min(final$compare))
+  finalarranged=as.data.frame(arrange(final,desc(compare)))
+  #success=as.numeric(table(finalarranged$`Outlier status`[1:N])[2])
+  #RESULT <- list("anomaly_indices" = listofoutliers, "number_of_true_detection" = success)
+  return(finalarranged)}
\ No newline at end of file