test.R

In Zelazny7/isofor: Isolation Forest Anomaly Detection

library(isofor)
library(microbenchmark)



set.seed(100)
mod1 <- iForest(iris, phi = 16, ncolsample = NULL, multicore = TRUE)

set.seed(100)
mod2 <- iForest(iris, ncolsample = 20, multicore=TRUE, nt=500)

p1 <- predict(mod1, titanic)
p2 <- predict(mod2, titanic)


code <- isofor_to_sas(mod, pfx = "test")

writeLines(code, "test.sas", sep = "\n")


p1 <- predict(mod, titanic, n.cores=1L)
p2 <- predict(mod, titanic, n.cores=4L)


microbenchmark(
  old = predict(mod, titanic, n.cores=1),
  new = predict(mod, titanic, n.cores=4), times=5L)

x <- titanic

for (i in 1:10) x <- rbind(x, x)

mod <- iForest(x, 500, 64, multicore = TRUE)
p1 <- predict(mod, x, n.cores = 4)


x <- mtcars

for (i in 1:3) x <- rbind(x, x)

microbenchmark(
  new=predict(mod, titanic, iterative = TRUE),
  old=predict(mod, titanic, iterative = FALSE), times = 5L)



sapply(mod$forest, function(t) {
  min(t[,"SplitValue"])
})