# nolint start
library(mlexperiments)
library(mllrnrs)# nolint start
library(mlexperiments)
library(mllrnrs)See https://github.com/kapsner/mllrnrs/blob/main/R/learner_xgboost.R for implementation details.
library(mlbench)
data("BreastCancer")
dataset <- BreastCancer |>
data.table::as.data.table() |>
na.omit()
feature_cols <- colnames(dataset)[2:10]
target_col <- "Class"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)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)]) - 1Lfold_list <- splitTools::create_folds(
y = train_y,
k = 3,
type = "stratified",
seed = seed
)# 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"
)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$execute(k = 3)
#>
#> Parameter settings [==================================================================>--------------------------------------------] 6/10 ( 60%)
#> Parameter settings [=============================================================================>---------------------------------] 7/10 ( 70%)
#> Parameter settings [========================================================================================>----------------------] 8/10 ( 80%)
#> Parameter settings [===================================================================================================>-----------] 9/10 ( 90%)
#> Parameter settings [==============================================================================================================] 10/10 (100%)
head(tuner_results_grid)
#> setting_id metric_optim_mean nrounds subsample colsample_bytree min_child_weight learning_rate max_depth objective eval_metric
#> <int> <num> <int> <num> <num> <num> <num> <num> <char> <char>
#> 1: 1 0.1473724 20 0.6 0.8 5 0.2 1 binary:logistic logloss
#> 2: 2 0.1836365 20 1.0 0.8 5 0.1 5 binary:logistic logloss
#> 3: 3 0.2004256 20 0.8 0.8 5 0.1 1 binary:logistic logloss
#> 4: 4 0.1431941 20 0.6 0.8 5 0.2 5 binary:logistic logloss
#> 5: 5 0.1766953 20 1.0 0.8 1 0.1 5 binary:logistic logloss
#> 6: 6 0.1866896 20 0.8 0.8 5 0.1 5 binary:logistic loglosstuner <- mlexperiments::MLTuneParameters$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)
#>
#> Number of rows of initialization grid > than 'options("mlexperiments.bayesian.max_init")'...
#> ... reducing initialization grid to 4 rows.
#> elapsed = 0.045 Round = 1 subsample = 0.8000 colsample_bytree = 0.8000 min_child_weight = 5.0000 learning_rate = 0.1000 max_depth = 1.0000 Value = -0.2004256
#> elapsed = 0.037 Round = 2 subsample = 0.6000 colsample_bytree = 1.0000 min_child_weight = 1.0000 learning_rate = 0.2000 max_depth = 1.0000 Value = -0.146329
#> elapsed = 0.033 Round = 3 subsample = 1.0000 colsample_bytree = 0.8000 min_child_weight = 5.0000 learning_rate = 0.1000 max_depth = 5.0000 Value = -0.1836365
#> elapsed = 0.033 Round = 4 subsample = 0.6000 colsample_bytree = 1.0000 min_child_weight = 5.0000 learning_rate = 0.2000 max_depth = 5.0000 Value = -0.143957
#> elapsed = 0.026 Round = 5 subsample = 0.2000 colsample_bytree = 0.2253119 min_child_weight = 6.0000 learning_rate = 0.1478213 max_depth = 9.0000 Value = -0.2851956
#> elapsed = 0.048 Round = 6 subsample = 1.0000 colsample_bytree = 0.6428195 min_child_weight = 7.0000 learning_rate = 0.2000 max_depth = 10.0000 Value = -0.1407496
#> elapsed = 0.029 Round = 7 subsample = 0.6370911 colsample_bytree = 0.8772461 min_child_weight = 3.0000 learning_rate = 0.1746002 max_depth = 4.0000 Value = -0.1423412
#> elapsed = 0.034 Round = 8 subsample = 1.0000 colsample_bytree = 1.0000 min_child_weight = 1.0000 learning_rate = 0.1753007 max_depth = 10.0000 Value = -0.1424366
#>
#> Best Parameters Found:
#> Round = 6 subsample = 1.0000 colsample_bytree = 0.6428195 min_child_weight = 7.0000 learning_rate = 0.2000 max_depth = 10.0000 Value = -0.1407496
head(tuner_results_bayesian)
#> setting_id subsample colsample_bytree min_child_weight learning_rate max_depth Value objective eval_metric metric_optim_mean
#> <int> <num> <num> <num> <num> <num> <num> <char> <char> <num>
#> 1: 1 0.8 0.8000000 5 0.1000000 1 -0.2004256 binary:logistic logloss 0.2004256
#> 2: 2 0.6 1.0000000 1 0.2000000 1 -0.1463290 binary:logistic logloss 0.1463290
#> 3: 3 1.0 0.8000000 5 0.1000000 5 -0.1836365 binary:logistic logloss 0.1836365
#> 4: 4 0.6 1.0000000 5 0.2000000 5 -0.1439570 binary:logistic logloss 0.1439570
#> 5: 5 0.2 0.2253119 6 0.1478213 9 -0.2851956 binary:logistic logloss 0.2851956
#> 6: 6 1.0 0.6428195 7 0.2000000 10 -0.1407496 binary:logistic logloss 0.1407496validator <- 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$execute()
#>
#> CV fold: Fold1
#>
#> CV fold: Fold2
#>
#> CV fold: Fold3
head(validator_results)
#> fold performance colsample_bytree min_child_weight learning_rate max_depth objective eval_metric nrounds
#> <char> <num> <num> <num> <num> <num> <char> <char> <int>
#> 1: Fold1 0.9920565 0.6428195 7 0.2 10 binary:logistic logloss 20
#> 2: Fold2 0.9928918 0.6428195 7 0.2 10 binary:logistic logloss 20
#> 3: Fold3 0.9818853 0.6428195 7 0.2 10 binary:logistic logloss 20validator <- mlexperiments::MLNestedCV$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 [=======================================================>-------------------------------------------------------] 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%)
#> CV fold: Fold2
#> CV progress [==============================================================================>----------------------------------------] 2/3 ( 67%)
#>
#> 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%)
#> CV fold: Fold3
#> CV progress [=======================================================================================================================] 3/3 (100%)
#>
#> 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 0.9876434 20 0.6 1 1 0.2 1 binary:logistic logloss
#> 2: Fold2 0.9905513 20 0.6 1 1 0.2 1 binary:logistic logloss
#> 3: Fold3 0.9843750 20 0.6 1 1 0.2 1 binary:logistic loglossvalidator <- 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
#>
#> Number of rows of initialization grid > than 'options("mlexperiments.bayesian.max_init")'...
#> ... reducing initialization grid to 4 rows.
#> elapsed = 0.045 Round = 1 subsample = 0.8000 colsample_bytree = 0.8000 min_child_weight = 5.0000 learning_rate = 0.1000 max_depth = 1.0000 Value = -0.2144507
#> elapsed = 0.025 Round = 2 subsample = 0.6000 colsample_bytree = 1.0000 min_child_weight = 1.0000 learning_rate = 0.2000 max_depth = 1.0000 Value = -0.1632438
#> elapsed = 0.028 Round = 3 subsample = 1.0000 colsample_bytree = 0.8000 min_child_weight = 5.0000 learning_rate = 0.1000 max_depth = 5.0000 Value = -0.1995263
#> elapsed = 0.037 Round = 4 subsample = 0.6000 colsample_bytree = 1.0000 min_child_weight = 5.0000 learning_rate = 0.2000 max_depth = 5.0000 Value = -0.1765465
#> elapsed = 0.052 Round = 5 subsample = 0.6587262 colsample_bytree = 0.2441275 min_child_weight = 4.0000 learning_rate = 0.1583349 max_depth = 8.0000 Value = -0.1751771
#> elapsed = 0.032 Round = 6 subsample = 0.2000 colsample_bytree = 0.7380485 min_child_weight = 1.0000 learning_rate = 0.1221712 max_depth = 6.0000 Value = -0.2001825
#> elapsed = 0.026 Round = 7 subsample = 0.4372448 colsample_bytree = 0.304417 min_child_weight = 1.0000 learning_rate = 0.1949787 max_depth = 8.0000 Value = -0.1653602
#> elapsed = 0.029 Round = 8 subsample = 1.0000 colsample_bytree = 0.3752773 min_child_weight = 1.0000 learning_rate = 0.176386 max_depth = 9.0000 Value = -0.151424
#>
#> Best Parameters Found:
#> Round = 8 subsample = 1.0000 colsample_bytree = 0.3752773 min_child_weight = 1.0000 learning_rate = 0.176386 max_depth = 9.0000 Value = -0.151424
#>
#> CV fold: Fold2
#> CV progress [==============================================================================>----------------------------------------] 2/3 ( 67%)
#>
#> Number of rows of initialization grid > than 'options("mlexperiments.bayesian.max_init")'...
#> ... reducing initialization grid to 4 rows.
#> elapsed = 0.024 Round = 1 subsample = 0.8000 colsample_bytree = 0.8000 min_child_weight = 5.0000 learning_rate = 0.1000 max_depth = 1.0000 Value = -0.2016618
#> elapsed = 0.024 Round = 2 subsample = 0.6000 colsample_bytree = 1.0000 min_child_weight = 1.0000 learning_rate = 0.2000 max_depth = 1.0000 Value = -0.1358879
#> elapsed = 0.036 Round = 3 subsample = 1.0000 colsample_bytree = 0.8000 min_child_weight = 5.0000 learning_rate = 0.1000 max_depth = 5.0000 Value = -0.1937136
#> elapsed = 0.025 Round = 4 subsample = 0.6000 colsample_bytree = 1.0000 min_child_weight = 5.0000 learning_rate = 0.2000 max_depth = 5.0000 Value = -0.1629105
#> elapsed = 0.026 Round = 5 subsample = 0.9597928 colsample_bytree = 0.2087203 min_child_weight = 1.0000 learning_rate = 0.1807996 max_depth = 3.0000 Value = -0.1409365
#> elapsed = 0.026 Round = 6 subsample = 0.9648661 colsample_bytree = 0.2000 min_child_weight = 10.0000 learning_rate = 0.2000 max_depth = 10.0000 Value = -0.1876724
#> elapsed = 0.027 Round = 7 subsample = 0.2232804 colsample_bytree = 0.89331 min_child_weight = 1.0000 learning_rate = 0.2000 max_depth = 10.0000 Value = -0.147581
#> elapsed = 0.051 Round = 8 subsample = 0.2000 colsample_bytree = 0.8707313 min_child_weight = 1.0000 learning_rate = 0.1000 max_depth = 10.0000 Value = -0.2066387
#>
#> Best Parameters Found:
#> Round = 2 subsample = 0.6000 colsample_bytree = 1.0000 min_child_weight = 1.0000 learning_rate = 0.2000 max_depth = 1.0000 Value = -0.1358879
#>
#> CV fold: Fold3
#> CV progress [=======================================================================================================================] 3/3 (100%)
#>
#> Number of rows of initialization grid > than 'options("mlexperiments.bayesian.max_init")'...
#> ... reducing initialization grid to 4 rows.
#> elapsed = 0.042 Round = 1 subsample = 0.8000 colsample_bytree = 0.8000 min_child_weight = 5.0000 learning_rate = 0.1000 max_depth = 1.0000 Value = -0.1996737
#> elapsed = 0.023 Round = 2 subsample = 0.6000 colsample_bytree = 1.0000 min_child_weight = 1.0000 learning_rate = 0.2000 max_depth = 1.0000 Value = -0.1550507
#> elapsed = 0.03 Round = 3 subsample = 1.0000 colsample_bytree = 0.8000 min_child_weight = 5.0000 learning_rate = 0.1000 max_depth = 5.0000 Value = -0.1921469
#> elapsed = 0.024 Round = 4 subsample = 0.6000 colsample_bytree = 1.0000 min_child_weight = 5.0000 learning_rate = 0.2000 max_depth = 5.0000 Value = -0.1630041
#> elapsed = 0.028 Round = 5 subsample = 0.8696134 colsample_bytree = 0.5493993 min_child_weight = 2.0000 learning_rate = 0.1541663 max_depth = 4.0000 Value = -0.156581
#> elapsed = 0.012 Round = 6 subsample = 0.2000 colsample_bytree = 0.4853539 min_child_weight = 10.0000 learning_rate = 0.1531602 max_depth = 10.0000 Value = -0.6478907
#> elapsed = 0.05 Round = 7 subsample = 0.2232806 colsample_bytree = 1.0000 min_child_weight = 3.0000 learning_rate = 0.1040753 max_depth = 10.0000 Value = -0.2244361
#> elapsed = 0.024 Round = 8 subsample = 0.5157127 colsample_bytree = 0.9291875 min_child_weight = 7.0000 learning_rate = 0.1998166 max_depth = 3.0000 Value = -0.2151049
#>
#> Best Parameters Found:
#> Round = 2 subsample = 0.6000 colsample_bytree = 1.0000 min_child_weight = 1.0000 learning_rate = 0.2000 max_depth = 1.0000 Value = -0.1550507
head(validator_results)
#> fold performance subsample colsample_bytree min_child_weight learning_rate max_depth objective eval_metric nrounds
#> <char> <num> <num> <num> <num> <num> <num> <char> <char> <int>
#> 1: Fold1 0.9955869 1.0 0.3752773 1 0.176386 9 binary:logistic logloss 20
#> 2: Fold2 0.9905513 0.6 1.0000000 1 0.200000 1 binary:logistic logloss 20
#> 3: Fold3 0.9843750 0.6 1.0000000 1 0.200000 1 binary:logistic logloss 20preds_xgboost <- mlexperiments::predictions(
object = validator,
newdata = test_x
)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 PPV
#> <char> <num> <num> <num> <num> <num> <int> <int> <int> <int> <num> <num> <num> <num> <num>
#> 1: Fold1 0.9839863 0.9839863 0.04247381 0.8131821 0.9369818 64 132 2 8 0.8888889 0.9850746 0.014925373 0.1111111 0.9696970
#> 2: Fold2 0.9812915 0.9812915 0.04675659 0.7943447 0.9059909 59 133 1 13 0.8194444 0.9925373 0.007462687 0.1805556 0.9833333
#> 3: Fold3 0.9830535 0.9830535 0.04884222 0.7851712 0.9161484 61 132 2 11 0.8472222 0.9850746 0.014925373 0.1527778 0.9682540
#> NPV FDR MCC F1 GMEAN GPR ACC MMCE BER
#> <num> <num> <num> <num> <num> <num> <num> <num> <num>
#> 1: 0.9428571 0.03030303 0.8930504 0.9275362 0.9357467 0.9284142 0.9514563 0.04854369 0.06301824
#> 2: 0.9109589 0.01666667 0.8521438 0.8939394 0.9018477 0.8976564 0.9320388 0.06796117 0.09400912
#> 3: 0.9230769 0.03174603 0.8613082 0.9037037 0.9135519 0.9057187 0.9368932 0.06310680 0.08385158