ledell

# Train 4 models and ensemble them together with new h2o.stack function.

# Requirements: Models must be same type of model H2OBinomial, etc
# Must have same outcome
# Must have used `fold_assignment = "Modulo"` and same number for `nfolds`, 
# or identical `fold_column` must be used to guarantee same folds between base models

# Requires: cvpreds.R and stack.R
source("https://gist.githubusercontent.com/ledell/f3a87bd136ce06e0a5ff/raw/2a82535892ff66694a1a401de46b8b5a92820849/cvpreds.R")
source("https://gist.githubusercontent.com/ledell/f389ac1e9c6e7000b299/raw/6bc1d2c9cfe1a51ffcdcf79cf184e80a40d4828f/stack.R")

ledell

# Extract cross-validated predicted values (in order of original rows)
h2o.cvpreds <- function(object, single_col = TRUE) {
  
  # TO DO: Check that object is an H2OModel
  # TO DO: Check that keep_cross_validation_predictions = TRUE in the model
  # TO DO: Need to add support for returning a multiclass prediction and binary (full frame: predict, p0, p1)
  # TO DO: Remove family variable and just check class(object) directly
  
  # Need to extract family from model object
  if (class(object) == "H2OBinomialModel") family <- "binomial"

ledell

# NOTE: This is now part of h2oEnsemble R package and should be used from there instead

# Given a list of H2O models, ensemble the base learners usig a metalearner (Stacking / Super Learning)

# Currently requires:
#source("cvpreds.R")


h2o.stack <- function(models,  #list of H2OModels
                      metalearner = "h2o.glm.wrapper", 

ledell

# The following two commands remove any previously installed H2O packages for R.
if ("package:h2o" %in% search()) { detach("package:h2o", unload=TRUE) }
if ("h2o" %in% rownames(installed.packages())) { remove.packages("h2o") }

# Next, we download packages that H2O depends on.
if (! ("methods" %in% rownames(installed.packages()))) { install.packages("methods") }
if (! ("statmod" %in% rownames(installed.packages()))) { install.packages("statmod") }
if (! ("stats" %in% rownames(installed.packages()))) { install.packages("stats") }
if (! ("graphics" %in% rownames(installed.packages()))) { install.packages("graphics") }
if (! ("RCurl" %in% rownames(installed.packages()))) { install.packages("RCurl") }

ledell

# install.packages("h2o", type="source", repos=(c("http://h2o-release.s3.amazonaws.com/h2o/rel-slater/5/R")))

library(h2o)
localH2O <- h2o.init(ip = "XX.XX.XX.XX", port = 54321)

#higgs <- h2o.importFile("/home/0xdiag/datasets/higgs/HIGGS.csv", destination_frame = "higgs")  #Local copy
higgs <- h2o.importFile("http://archive.ics.uci.edu/ml/machine-learning-databases/00280/HIGGS.csv.gz", destination_frame = "higgs")
dim(higgs)  #11M x 29

higgs$C1 <- as.factor(higgs$C1)  #Encode response as categorical

ledell

### Lending Club example using cleaned up dataset & h2o.ensemble ###

library(h2o)
h2o.init(nthreads = -1, max_mem_size = "8G")

loan_csv <- "https://raw.githubusercontent.com/h2oai/app-consumer-loan/master/data/loan.csv"

data <- h2o.importFile(loan_csv)  # 163994 x 15
data$bad_loan <- as.factor(data$bad_loan)
rand  <- h2o.runif(data, seed = 1)

ledell

library(h2o)
h2o.init(nthreads = -1)  # This means nthreads = num available cores

train_file <- "https://h2o-public-test-data.s3.amazonaws.com/bigdata/laptop/mnist/train.csv.gz"
test_file <- "https://h2o-public-test-data.s3.amazonaws.com/bigdata/laptop/mnist/test.csv.gz"

train <- h2o.importFile(train_file)
test <- h2o.importFile(test_file)

# To see a brief summary of the data, run the following command

ledell

# Example of how to train an H2O model with folds that are 
# stratified both by outcome and a cluster id

library(cvAUC)

data("adherence")  #load a dataset with an ID column
df <- adherence

# Load a utility function for creating stratified folds
source("https://gist.githubusercontent.com/ledell/bd4e227d4e5ff426c41d/raw/708eb429fa1954a140d65a6a42ce93847affd67c/CVFolds2.R")  #utility function

ledell

# Update for SuperLearner::CVFolds function that enables stratification by outcome and cluster ID

CVFolds2 <- function (N, id, Y, cvControl) {
    if (!is.null(cvControl$validRows)) {
        return(cvControl$validRows)
    }
    stratifyCV <- cvControl$stratifyCV
    shuffle <- cvControl$shuffle
    V <- cvControl$V
    if (!stratifyCV) {  ### Not Stratified

ledell

library(devtools)
install_github("h2oai/h2o-3/h2o-r/ensemble/h2oEnsemble-package")