4-h2o.R

library(h2o)

h2oServer <- h2o.init(max_mem_size="60g", nthreads=-1)


dx_train <- h2o.importFile(h2oServer, path = "train-0.1m.csv")
dx_test <- h2o.importFile(h2oServer, path = "test.csv")


Xnames <- names(dx_train)[which(names(dx_train)!="dep_delayed_15min")]

system.time({
  md <- h2o.randomForest(x = Xnames, y = "dep_delayed_15min", data = dx_train, ntree = 500, type="BigData")
})


system.time({
  phat <- h2o.predict(md, dx_test)[,"Y"]
  print(h2o.performance(phat, dx_test[,"dep_delayed_15min"])@model$auc)
})