xgboost: Binary Classification

# nolint start
library(mlexperiments)
library(mllrnrs)

See https://github.com/kapsner/mllrnrs/blob/main/R/learner_xgboost.R for implementation details.

Preprocessing

Import and Prepare Data

library(mlbench)
data("PimaIndiansDiabetes2")
dataset <- PimaIndiansDiabetes2 |>
  data.table::as.data.table() |>
  na.omit()

feature_cols <- colnames(dataset)[1:8]
target_col <- "diabetes"

General Configurations

seed <- 123
if (isTRUE(as.logical(Sys.getenv("_R_CHECK_LIMIT_CORES_")))) {
  # on cran
  ncores <- 2L
} else {
  ncores <- ifelse(
    test = parallel::detectCores() > 4,
    yes = 4L,
    no = ifelse(
      test = parallel::detectCores() < 2L,
      yes = 1L,
      no = parallel::detectCores()
    )
  )
}
options("mlexperiments.bayesian.max_init" = 4L)
options("mlexperiments.optim.xgb.nrounds" = 20L)
options("mlexperiments.optim.xgb.early_stopping_rounds" = 5L)

Generate Training- and Test Data

data_split <- splitTools::partition(
  y = dataset[, get(target_col)],
  p = c(train = 0.7, test = 0.3),
  type = "stratified",
  seed = seed
)

train_x <- model.matrix(
  ~ -1 + .,
  dataset[data_split$train, .SD, .SDcols = feature_cols]
)
train_y <- as.integer(dataset[data_split$train, get(target_col)]) - 1L


test_x <- model.matrix(
  ~ -1 + .,
  dataset[data_split$test, .SD, .SDcols = feature_cols]
)
test_y <- as.integer(dataset[data_split$test, get(target_col)]) - 1L

Generate Training Data Folds

fold_list <- splitTools::create_folds(
  y = train_y,
  k = 3,
  type = "stratified",
  seed = seed
)

Experiments

Prepare Experiments

# required learner arguments, not optimized
learner_args <- list(
  objective = "binary:logistic",
  eval_metric = "logloss"
)

# set arguments for predict function and performance metric,
# required for mlexperiments::MLCrossValidation and
# mlexperiments::MLNestedCV
predict_args <- NULL
performance_metric <- metric("auc")
performance_metric_args <- list(positive = "1", negative = "0")
return_models <- FALSE

# required for grid search and initialization of bayesian optimization
parameter_grid <- expand.grid(
  subsample = seq(0.6, 1, .2),
  colsample_bytree = seq(0.6, 1, .2),
  min_child_weight = seq(1, 5, 4),
  learning_rate = seq(0.1, 0.2, 0.1),
  max_depth = seq(1, 5, 4)
)
# reduce to a maximum of 10 rows
if (nrow(parameter_grid) > 10) {
  set.seed(123)
  sample_rows <- sample(seq_len(nrow(parameter_grid)), 10, FALSE)
  parameter_grid <- kdry::mlh_subset(parameter_grid, sample_rows)
}

# required for bayesian optimization
parameter_bounds <- list(
  subsample = c(0.2, 1),
  colsample_bytree = c(0.2, 1),
  min_child_weight = c(1L, 10L),
  learning_rate = c(0.1, 0.2),
  max_depth =  c(1L, 10L)
)
optim_args <- list(
  n_iter = ncores,
  kappa = 3.5,
  acq = "ucb"
)

Hyperparameter Tuning

tuner <- mlexperiments::MLTuneParameters$new(
  learner = mllrnrs::LearnerXgboost$new(
    metric_optimization_higher_better = FALSE
  ),
  strategy = "grid",
  ncores = ncores,
  seed = seed
)

tuner$parameter_grid <- parameter_grid
tuner$learner_args <- learner_args
tuner$split_type <- "stratified"

tuner$set_data(
  x = train_x,
  y = train_y
)

tuner_results_grid <- tuner$execute(k = 3)
#>
#> Parameter settings [==================>-----------------------------------------------------------------------------] 2/10 ( 20%)
#> Parameter settings [============================>-------------------------------------------------------------------] 3/10 ( 30%)
#> Parameter settings [=====================================>----------------------------------------------------------] 4/10 ( 40%)
#> Parameter settings [===============================================>------------------------------------------------] 5/10 ( 50%)
#> Parameter settings [=========================================================>--------------------------------------] 6/10 ( 60%)
#> Parameter settings [==================================================================>-----------------------------] 7/10 ( 70%)
#> Parameter settings [============================================================================>-------------------] 8/10 ( 80%)
#> Parameter settings [=====================================================================================>----------] 9/10 ( 90%)
#> Parameter settings [===============================================================================================] 10/10 (100%)

head(tuner_results_grid)
#>    setting_id metric_optim_mean nrounds subsample colsample_bytree min_child_weight learning_rate max_depth       objective eval_metric
#>         <int>             <num>   <int>     <num>            <num>            <num>         <num>     <num>          <char>      <char>
#> 1:          1         0.3977489      62       0.6              0.8                5           0.2         1 binary:logistic     logloss
#> 2:          2         0.3915203      67       1.0              0.8                5           0.1         5 binary:logistic     logloss
#> 3:          3         0.3972711      96       0.8              0.8                5           0.1         1 binary:logistic     logloss
#> 4:          4         0.3951791      62       0.6              0.8                5           0.2         5 binary:logistic     logloss
#> 5:          5         0.3786375      44       1.0              0.8                1           0.1         5 binary:logistic     logloss
#> 6:          6         0.3956902      75       0.8              0.8                5           0.1         5 binary:logistic     logloss

Bayesian Optimization

tuner <- mlexperiments::MLTuneParameters$new(
  learner = mllrnrs::LearnerXgboost$new(
    metric_optimization_higher_better = FALSE
  ),
  strategy = "bayesian",
  ncores = ncores,
  seed = seed
)

tuner$parameter_grid <- parameter_grid
tuner$parameter_bounds <- parameter_bounds

tuner$learner_args <- learner_args
tuner$optim_args <- optim_args

tuner$split_type <- "stratified"

tuner$set_data(
  x = train_x,
  y = train_y
)

tuner_results_bayesian <- tuner$execute(k = 3)
#>
#> Registering parallel backend using 4 cores.

head(tuner_results_bayesian)
#>    Epoch setting_id subsample colsample_bytree min_child_weight learning_rate max_depth gpUtility acqOptimum inBounds Elapsed
#>    <num>      <int>     <num>            <num>            <num>         <num>     <num>     <num>     <lgcl>   <lgcl>   <num>
#> 1:     0          1       0.6              0.8                5           0.2         1        NA      FALSE     TRUE   0.890
#> 2:     0          2       1.0              0.8                5           0.1         5        NA      FALSE     TRUE   0.892
#> 3:     0          3       0.8              0.8                5           0.1         1        NA      FALSE     TRUE   0.991
#> 4:     0          4       0.6              0.8                5           0.2         5        NA      FALSE     TRUE   0.911
#> 5:     0          5       1.0              0.8                1           0.1         5        NA      FALSE     TRUE   0.173
#> 6:     0          6       0.8              0.8                5           0.1         5        NA      FALSE     TRUE   0.183
#>         Score metric_optim_mean nrounds errorMessage       objective eval_metric
#>         <num>             <num>   <int>       <lgcl>          <char>      <char>
#> 1: -0.3977489         0.3977489      62           NA binary:logistic     logloss
#> 2: -0.3915203         0.3915203      67           NA binary:logistic     logloss
#> 3: -0.3972711         0.3972711      96           NA binary:logistic     logloss
#> 4: -0.3951791         0.3951791      62           NA binary:logistic     logloss
#> 5: -0.3786375         0.3786375      44           NA binary:logistic     logloss
#> 6: -0.3956902         0.3956902      75           NA binary:logistic     logloss

k-Fold Cross Validation

validator <- mlexperiments::MLCrossValidation$new(
  learner = mllrnrs::LearnerXgboost$new(
    metric_optimization_higher_better = FALSE
  ),
  fold_list = fold_list,
  ncores = ncores,
  seed = seed
)

validator$learner_args <- tuner$results$best.setting[-1]

validator$predict_args <- predict_args
validator$performance_metric <- performance_metric
validator$performance_metric_args <- performance_metric_args
validator$return_models <- return_models

validator$set_data(
  x = train_x,
  y = train_y
)

validator_results <- validator$execute()
#>
#> CV fold: Fold1
#>
#> CV fold: Fold2
#>
#> CV fold: Fold3

head(validator_results)
#>      fold performance subsample colsample_bytree min_child_weight learning_rate max_depth nrounds       objective eval_metric
#>    <char>       <num>     <num>            <num>            <num>         <num>     <num>   <int>          <char>      <char>
#> 1:  Fold1   0.8947647         1              0.8                1           0.1         5      44 binary:logistic     logloss
#> 2:  Fold2   0.8720254         1              0.8                1           0.1         5      44 binary:logistic     logloss
#> 3:  Fold3   0.9010741         1              0.8                1           0.1         5      44 binary:logistic     logloss

Nested Cross Validation

validator <- mlexperiments::MLNestedCV$new(
  learner = mllrnrs::LearnerXgboost$new(
    metric_optimization_higher_better = FALSE
  ),
  strategy = "grid",
  fold_list = fold_list,
  k_tuning = 3L,
  ncores = ncores,
  seed = seed
)

validator$parameter_grid <- parameter_grid
validator$learner_args <- learner_args
validator$split_type <- "stratified"

validator$predict_args <- predict_args
validator$performance_metric <- performance_metric
validator$performance_metric_args <- performance_metric_args
validator$return_models <- return_models

validator$set_data(
  x = train_x,
  y = train_y
)

validator_results <- validator$execute()
#>
#> CV fold: Fold1
#>
#> Parameter settings [==================================================================>-----------------------------] 7/10 ( 70%)
#> Parameter settings [============================================================================>-------------------] 8/10 ( 80%)
#> Parameter settings [=====================================================================================>----------] 9/10 ( 90%)
#> Parameter settings [===============================================================================================] 10/10 (100%)
#> CV fold: Fold2
#> CV progress [====================================================================>-----------------------------------] 2/3 ( 67%)
#>
#> Parameter settings [==================================================================>-----------------------------] 7/10 ( 70%)
#> Parameter settings [============================================================================>-------------------] 8/10 ( 80%)
#> Parameter settings [=====================================================================================>----------] 9/10 ( 90%)
#> Parameter settings [===============================================================================================] 10/10 (100%)
#> CV fold: Fold3
#> CV progress [========================================================================================================] 3/3 (100%)
#>
#> Parameter settings [===============================================>------------------------------------------------] 5/10 ( 50%)
#> Parameter settings [=========================================================>--------------------------------------] 6/10 ( 60%)
#> Parameter settings [==================================================================>-----------------------------] 7/10 ( 70%)
#> Parameter settings [============================================================================>-------------------] 8/10 ( 80%)
#> Parameter settings [=====================================================================================>----------] 9/10 ( 90%)
#> Parameter settings [===============================================================================================] 10/10 (100%)

head(validator_results)
#>      fold performance nrounds subsample colsample_bytree min_child_weight learning_rate max_depth       objective eval_metric
#>    <char>       <num>   <int>     <num>            <num>            <num>         <num>     <num>          <char>      <char>
#> 1:  Fold1   0.8714966      40       0.6              1.0                1           0.2         1 binary:logistic     logloss
#> 2:  Fold2   0.8754627      35       1.0              1.0                5           0.1         5 binary:logistic     logloss
#> 3:  Fold3   0.8883550      41       0.8              0.8                5           0.1         1 binary:logistic     logloss

Inner Bayesian Optimization

validator <- mlexperiments::MLNestedCV$new(
  learner = mllrnrs::LearnerXgboost$new(
    metric_optimization_higher_better = FALSE
  ),
  strategy = "bayesian",
  fold_list = fold_list,
  k_tuning = 3L,
  ncores = ncores,
  seed = seed
)

validator$parameter_grid <- parameter_grid
validator$learner_args <- learner_args
validator$split_type <- "stratified"


validator$parameter_bounds <- parameter_bounds
validator$optim_args <- optim_args

validator$predict_args <- predict_args
validator$performance_metric <- performance_metric
validator$performance_metric_args <- performance_metric_args
validator$return_models <- TRUE

validator$set_data(
  x = train_x,
  y = train_y
)

validator_results <- validator$execute()
#>
#> CV fold: Fold1
#>
#> Registering parallel backend using 4 cores.
#>
#> CV fold: Fold2
#> CV progress [====================================================================>-----------------------------------] 2/3 ( 67%)
#>
#> Registering parallel backend using 4 cores.
#>
#> CV fold: Fold3
#> CV progress [========================================================================================================] 3/3 (100%)
#>
#> Registering parallel backend using 4 cores.

head(validator_results)
#>      fold performance subsample colsample_bytree min_child_weight learning_rate max_depth nrounds       objective eval_metric
#>    <char>       <num>     <num>            <num>            <num>         <num>     <num>   <int>          <char>      <char>
#> 1:  Fold1   0.8714966       0.6        1.0000000                1     0.2000000         1      40 binary:logistic     logloss
#> 2:  Fold2   0.8754627       1.0        1.0000000                5     0.1000000         5      35 binary:logistic     logloss
#> 3:  Fold3   0.8810062       1.0        0.6293304                1     0.1034034         1      56 binary:logistic     logloss

Holdout Test Dataset Performance

Predict Outcome in Holdout Test Dataset

preds_xgboost <- mlexperiments::predictions(
  object = validator,
  newdata = test_x
)

Evaluate Performance on Holdout Test Dataset

perf_xgboost <- mlexperiments::performance(
  object = validator,
  prediction_results = preds_xgboost,
  y_ground_truth = test_y,
  type = "binary"
)
perf_xgboost
#>     model performance       AUC     Brier BrierScaled       BAC    TP    TN    FP    FN       TPR       TNR       FPR       FNR
#>    <char>       <num>     <num>     <num>       <num>     <num> <int> <int> <int> <int>     <num>     <num>     <num>     <num>
#> 1:  Fold1   0.7913015 0.7913015 0.1743251   0.2121706 0.6994482    20    70     9    19 0.5128205 0.8860759 0.1139241 0.4871795
#> 2:  Fold2   0.7745862 0.7745862 0.1856610   0.1609401 0.6481662    16    70     9    23 0.4102564 0.8860759 0.1139241 0.5897436
#> 3:  Fold3   0.7917884 0.7917884 0.1739823   0.2137198 0.6609867    17    70     9    22 0.4358974 0.8860759 0.1139241 0.5641026
#>          PPV       NPV       FDR       MCC        F1     GMEAN       GPR       ACC      MMCE       BER
#>        <num>     <num>     <num>     <num>     <num>     <num>     <num>     <num>     <num>     <num>
#> 1: 0.6896552 0.7865169 0.3103448 0.4358249 0.5882353 0.6740904 0.5947010 0.7627119 0.2372881 0.3005518
#> 2: 0.6400000 0.7526882 0.3600000 0.3411249 0.5000000 0.6029248 0.5124101 0.7288136 0.2711864 0.3518338
#> 3: 0.6538462 0.7608696 0.3461538 0.3654140 0.5230769 0.6214807 0.5338631 0.7372881 0.2627119 0.3390133