R/LearnerClassifCatboost.R
In mlr3learners/mlr3learners.catboost: Learners from catboost package for mlr3
#' @title Gradient Boosted Decision Trees Classification Learner
#'
#' @name mlr_learners_classif.catboost
#'
#' @description
#' Gradient boosted decision trees classification learner.
#' Uses [catboost::catboost.train()] and [catboost::catboost.predict()] from
#' package catboost.
#'
#' By default, the log loss is optimized as the loss function for binary
#' classification, and the multi-class loss is optimized as the loss function
#' for multiclass classification.
#'
#' @section Custom mlr3 defaults:
#' - `logging_level`:
#' - Actual default: "Verbose"
#' - Adjusted default: "Silent"
#' - Reason for change: consistent with other mlr3 learners
#' - `thread_count`:
#' - Actual default: -1
#' - Adjusted default: 1
#' - Reason for change: consistent with other mlr3 learners
#' - `allow_writing_files`:
#' - Actual default: TRUE
#' - Adjusted default: FALSE
#' - Reason for change: consistent with other mlr3 learners
#' - `save_snapshot`:
#' - Actual default: TRUE
#' - Adjusted default: FALSE
#' - Reason for change: consistent with other mlr3 learners
#'
#' @templateVar id classif.catboost
#' @template section_dictionary_learner
#'
#' @references
#' \cite{mlr3learners.catboost}{dorogush2019}
#' \cite{mlr3learners.catboost}{prokhorenkova2017}
#'
#' @export
#' @template seealso_learner
#' @template example
LearnerClassifCatboost = R6Class("LearnerClassifCatboost",
inherit = LearnerClassif,
public = list(
#' @description
#' Create a `LearnerClassifCatboost` object.
initialize = function() {
ps = ParamSet$new(
params = list(
# catboost.train
# https://catboost.ai/docs/concepts/r-reference_catboost-train.html
# Common parameters
ParamFct$new(
id = "loss_function_twoclass",
levels = c("Logloss", "CrossEntropy"),
default = "Logloss", tags = "train"),
ParamFct$new(
id = "loss_function_multiclass",
levels = c("MultiClass", "MultiClassOneVsAll"),
default = "MultiClass", tags = "train"),
# custom_loss missing
# eval_metric missing
ParamInt$new(
id = "iterations", lower = 1L, upper = Inf,
default = 1000, tags = "train"),
ParamDbl$new(
id = "learning_rate", lower = 0.001, upper = 1,
default = 0.03, tags = "train"),
ParamInt$new(
id = "random_seed", lower = 0, upper = Inf,
default = 0, tags = "train"),
ParamDbl$new(
id = "l2_leaf_reg", lower = 0, upper = Inf,
default = 3, tags = "train"),
ParamFct$new(
id = "bootstrap_type",
levels = c("Bayesian", "Bernoulli", "MVS", "Poisson", "No"),
tags = "train"),
ParamDbl$new(
id = "bagging_temperature", lower = 0, upper = Inf,
default = 1, tags = "train"),
ParamDbl$new(id = "subsample", lower = 0, upper = 1, tags = "train"),
ParamFct$new(
id = "sampling_frequency",
levels = c("PerTree", "PerTreeLevel"),
default = "PerTreeLevel", tags = "train"),
ParamFct$new(
id = "sampling_unit", levels = c("Object", "Group"),
default = "Object", tags = "train"),
ParamDbl$new(id = "mvs_reg", lower = 0, upper = Inf, tags = "train"),
ParamDbl$new(
id = "random_strength", lower = 0, upper = Inf,
default = 1, tags = "train"),
# use_best_model missing
# best_model_min_trees missing
ParamInt$new(
id = "depth", lower = 1L, upper = 16L,
default = 6L, tags = "train"),
ParamFct$new(
id = "grow_policy",
levels = c("SymmetricTree", "Depthwise", "Lossguide"),
default = "SymmetricTree", tags = "train"),
ParamInt$new(
id = "min_data_in_leaf", lower = 1L, upper = Inf,
default = 1L, tags = "train"),
ParamInt$new(
id = "max_leaves", lower = 1L, upper = Inf,
default = 31L, tags = "train"),
# ignored_features missing
# one_hot_max_size missing
ParamLgl$new(id = "has_time", default = FALSE, tags = "train"),
ParamDbl$new(
id = "rsm", lower = 0.001, upper = 1,
default = 1, tags = "train"),
ParamFct$new(
id = "nan_mode", levels = c("Min", "Max"),
default = "Min", tags = "train"), # do not allow "Forbidden"
ParamInt$new(
id = "fold_permutation_block",
lower = 1L, upper = 256L, tags = "train"),
ParamFct$new(
id = "leaf_estimation_method",
levels = c("Newton", "Gradient", "Exact"), tags = "train"),
ParamInt$new(
id = "leaf_estimation_iterations",
lower = 1L, upper = Inf, tags = "train"),
ParamFct$new(
id = "leaf_estimation_backtracking",
levels = c("No", "AnyImprovement", "Armijo"),
default = "AnyImprovement", tags = "train"),
# name missing
ParamDbl$new(
id = "fold_len_multiplier",
lower = 1.001, upper = Inf, default = 2, tags = "train"),
ParamLgl$new(
id = "approx_on_full_history",
default = TRUE, tags = "train"),
ParamUty$new(id = "class_weights", tags = "train"),
ParamFct$new(
id = "auto_class_weights",
levels = c("None", "Balanced", "SqrtBalanced"),
default = "None", tags = "train"),
ParamFct$new(
id = "boosting_type",
levels = c("Ordered", "Plain"), tags = "train"),
ParamLgl$new("boost_from_average", tags = "train"),
ParamLgl$new(id = "langevin", default = FALSE, tags = "train"),
ParamDbl$new(
id = "diffusion_temperature", lower = 0, upper = Inf,
default = 10000, tags = "train"),
# allow_const_label missing
ParamFct$new(
id = "score_function",
levels = c("Cosine", "L2", "NewtonCosine", "NewtonL2"),
default = "Cosine", tags = "train"),
# cat_features missing
ParamUty$new(
id = "monotone_constraints", tags = "train",
custom_check = checkmate::check_string),
ParamUty$new(
id = "feature_weights", tags = "train",
custom_check = checkmate::check_string),
ParamUty$new(
id = "first_feature_use_penalties", tags = "train",
custom_check = checkmate::check_string),
ParamDbl$new(
id = "penalties_coefficient", lower = 0, upper = Inf,
default = 1, tags = "train"),
ParamUty$new(
id = "per_object_feature_penalties", tags = "train",
custom_check = checkmate::check_string),
ParamDbl$new(id = "model_shrink_rate", tags = "train"),
ParamFct$new(
id = "model_shrink_mode",
levels = c("Constant", "Decreasing"), tags = "train"),
# Overfitting detection settings missing
# Quantization settings
ParamDbl$new(id = "target_border", tags = "train"),
ParamInt$new(
id = "border_count",
lower = 1L, upper = 65535L, tags = "train"),
ParamFct$new(
id = "feature_border_type",
levels = c(
"Median", "Uniform", "UniformAndQuantiles", "MaxLogSum",
"MinEntropy", "GreedyLogSum"),
default = "GreedyLogSum", tags = "train"),
ParamUty$new(
id = "per_float_feature_quantization", tags = "train",
custom_check = checkmate::check_string),
# Multiclassification settings
ParamInt$new(
id = "classes_count",
lower = 1L, upper = Inf, tags = "train"),
# Performance Settings
ParamInt$new(
id = "thread_count",
lower = -1L, upper = Inf, default = 1L,
tags = c("train", "predict", "importance")),
# Processing units settings
ParamFct$new(
id = "task_type", levels = c("CPU", "GPU"),
default = "CPU", tags = "train"),
ParamUty$new(id = "devices", tags = "train"),
# Output settings
ParamFct$new(
id = "logging_level",
levels = c("Silent", "Verbose", "Info", "Debug"),
default = "Silent", tags = "train"),
ParamInt$new(
id = "metric_period", lower = 1L, upper = Inf,
default = 1L, tags = "train"),
# verbose missing
ParamUty$new(
id = "train_dir", default = "catboost_info",
tags = "train", custom_check = checkmate::check_string),
ParamDbl$new(
id = "model_size_reg", lower = 0, upper = 1,
default = 0.5, tags = "train"),
ParamLgl$new(
id = "allow_writing_files",
default = FALSE, tags = "train"),
ParamLgl$new(id = "save_snapshot", default = FALSE, tags = "train"),
ParamUty$new(
id = "snapshot_file", tags = "train",
custom_check = checkmate::check_string),
ParamInt$new(
id = "snapshot_interval", lower = 1L, upper = Inf,
default = 600L, tags = "train"),
# CTR settings
ParamUty$new(
id = "simple_ctr", tags = "train",
custom_check = checkmate::check_string),
ParamUty$new(
id = "combinations_ctr", tags = "train",
custom_check = checkmate::check_string),
ParamInt$new(
id = "ctr_target_border_count",
lower = 1L, upper = 255L, tags = "train"),
ParamFct$new(
id = "counter_calc_method",
levels = c("SkipTest", "Full"), default = "Full", tags = "train"),
ParamInt$new(
id = "max_ctr_complexity",
lower = 1L, upper = Inf, tags = "train"),
ParamInt$new(
id = "ctr_leaf_count_limit",
lower = 1L, upper = Inf, tags = "train"),
ParamLgl$new(
id = "store_all_simple_ctr",
default = FALSE, tags = "train"),
ParamFct$new(
id = "final_ctr_computation_mode",
levels = c("Default", "Skip"), default = "Default", tags = "train"),
# catboost.predict
# https://catboost.ai/docs/concepts/r-reference_catboost-predict.html
ParamLgl$new(id = "verbose", default = FALSE, tags = "predict"),
ParamInt$new(
id = "ntree_start", lower = 0L, upper = Inf,
default = 0L, tags = "predict"),
ParamInt$new(
id = "ntree_end", lower = 0L, upper = Inf,
default = 0L, tags = "predict")
)
)
ps$add_dep(
id = "mvs_reg", on = "bootstrap_type",
cond = CondEqual$new("MVS"))
ps$add_dep(
id = "min_data_in_leaf", on = "grow_policy",
cond = CondAnyOf$new(c("Depthwise", "Lossguide")))
ps$add_dep(
id = "diffusion_temperature", on = "langevin",
cond = CondEqual$new(TRUE))
ps$values$loss_function_twoclass = "Logloss"
ps$values$loss_function_multiclass = "MultiClass"
ps$values$logging_level = "Silent"
ps$values$thread_count = 1L
ps$values$allow_writing_files = FALSE
ps$values$save_snapshot = FALSE
super$initialize(
id = "classif.catboost",
packages = "catboost",
feature_types = c("logical", "integer", "numeric", "factor", "ordered"),
predict_types = c("response", "prob"),
param_set = ps,
properties = c(
"missings", "weights", "importance", "twoclass", "multiclass"), # FIXME: parallel
man = "mlr3learners.catboost::mlr_learners_classif.catboost"
)
},
#' @description
#' The importance scores are calculated using
#' [`catboost.get_feature_importance`][catboost::catboost.get_feature_importance],
#' setting `type = "FeatureImportance"`, returned for 'all'.
#' @return Named `numeric()`.
importance = function() {
# https://catboost.ai/docs/concepts/r-reference_catboost-get_feature_importance.html
imp = mlr3misc::invoke(catboost::catboost.get_feature_importance,
model = self$model,
type = "FeatureImportance",
thread_count = self$param_set$values$thread_count)
names(imp) = self$state$train_task$feature_names
sort(imp, decreasing = TRUE)
}
),
private = list(
.train = function(task) {
# integer/logical features must be converted to numerics explicitly
data = task$data(cols = task$feature_names)
to_numerics = task$feature_types$id[task$feature_types$type %in%
c("integer", "logical")]
if (length(to_numerics)) {
data[, (to_numerics) := lapply(.SD, as.numeric), .SDcols = to_numerics]
}
# target is encoded as integer values from 0
# if binary, the positive class is 1
is_binary = (length(task$class_names) == 2L)
label = if (is_binary) {
ifelse(task$data(cols = task$target_names)[[1L]] == task$positive,
yes = 1L,
no = 0L)
} else {
as.integer(task$data(cols = task$target_names)[[1L]]) - 1L
}
# data must be a dataframe
learn_pool = mlr3misc::invoke(catboost::catboost.load_pool,
data = setDF(data),
label = label,
weight = task$weights$weight,
thread_count = self$param_set$values$thread_count)
# set loss_function correctly
pars = self$param_set$get_values(tags = "train")
pars$loss_function = if (is_binary) {
pars$loss_function_twoclass
} else {
pars$loss_function_multiclass
}
pars$loss_function_twoclass = NULL
pars$loss_function_multiclass = NULL
mlr3misc::invoke(catboost::catboost.train,
learn_pool = learn_pool,
test_pool = NULL,
params = pars)
},
.predict = function(task) {
# integer/logical features must be converted to numerics explicitly
data = task$data(cols = task$feature_names)
to_numerics = task$feature_types$id[task$feature_types$type %in%
c("integer", "logical")]
if (length(to_numerics)) {
data[, (to_numerics) := lapply(.SD, as.numeric), .SDcols = to_numerics]
}
# target was encoded as integer values based on the train_task
# to later revert this, again use the train_task
is_binary = (length(self$state$train_task$class_names) == 2L)
# data must be a dataframe
pool = mlr3misc::invoke(catboost::catboost.load_pool,
data = setDF(data),
thread_count = self$param_set$values$thread_count)
prediction_type = if (self$predict_type == "response") {
"Class"
} else {
"Probability"
}
preds = mlr3misc::invoke(catboost::catboost.predict,
model = self$model,
pool = pool,
prediction_type = prediction_type,
.args = self$param_set$get_values(tags = "predict"))
if (self$predict_type == "response") {
response = if (is_binary) {
ifelse(preds == 1L,
yes = self$state$train_task$positive,
no = setdiff(
self$state$train_task$class_names,
self$state$train_task$positive))
} else {
self$state$train_task$class_names[preds + 1L]
}
PredictionClassif$new(task = task, response = response)
} else {
if (is_binary && is.null(dim(preds))) {
preds = matrix(c(preds, 1 - preds), ncol = 2L, nrow = length(preds))
}
colnames(preds) = self$state$train_task$class_names
PredictionClassif$new(task = task, prob = preds)
}
}
)
)
mlr3learners/mlr3learners.catboost documentation built on July 27, 2020, 7:30 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
#' @title Gradient Boosted Decision Trees Classification Learner
#'
#' @name mlr_learners_classif.catboost
#'
#' @description
#' Gradient boosted decision trees classification learner.
#' Uses [catboost::catboost.train()] and [catboost::catboost.predict()] from
#' package catboost.
#'
#' By default, the log loss is optimized as the loss function for binary
#' classification, and the multi-class loss is optimized as the loss function
#' for multiclass classification.
#'
#' @section Custom mlr3 defaults:
#' - `logging_level`:
#' - Actual default: "Verbose"
#' - Adjusted default: "Silent"
#' - Reason for change: consistent with other mlr3 learners
#' - `thread_count`:
#' - Actual default: -1
#' - Adjusted default: 1
#' - Reason for change: consistent with other mlr3 learners
#' - `allow_writing_files`:
#' - Actual default: TRUE
#' - Adjusted default: FALSE
#' - Reason for change: consistent with other mlr3 learners
#' - `save_snapshot`:
#' - Actual default: TRUE
#' - Adjusted default: FALSE
#' - Reason for change: consistent with other mlr3 learners
#'
#' @templateVar id classif.catboost
#' @template section_dictionary_learner
#'
#' @references
#' \cite{mlr3learners.catboost}{dorogush2019}
#' \cite{mlr3learners.catboost}{prokhorenkova2017}
#'
#' @export
#' @template seealso_learner
#' @template example
LearnerClassifCatboost = R6Class("LearnerClassifCatboost",
inherit = LearnerClassif,
public = list(
#' @description
#' Create a `LearnerClassifCatboost` object.
initialize = function() {
ps = ParamSet$new(
params = list(
# catboost.train
# https://catboost.ai/docs/concepts/r-reference_catboost-train.html
# Common parameters
ParamFct$new(
id = "loss_function_twoclass",
levels = c("Logloss", "CrossEntropy"),
default = "Logloss", tags = "train"),
ParamFct$new(
id = "loss_function_multiclass",
levels = c("MultiClass", "MultiClassOneVsAll"),
default = "MultiClass", tags = "train"),
# custom_loss missing
# eval_metric missing
ParamInt$new(
id = "iterations", lower = 1L, upper = Inf,
default = 1000, tags = "train"),
ParamDbl$new(
id = "learning_rate", lower = 0.001, upper = 1,
default = 0.03, tags = "train"),
ParamInt$new(
id = "random_seed", lower = 0, upper = Inf,
default = 0, tags = "train"),
ParamDbl$new(
id = "l2_leaf_reg", lower = 0, upper = Inf,
default = 3, tags = "train"),
ParamFct$new(
id = "bootstrap_type",
levels = c("Bayesian", "Bernoulli", "MVS", "Poisson", "No"),
tags = "train"),
ParamDbl$new(
id = "bagging_temperature", lower = 0, upper = Inf,
default = 1, tags = "train"),
ParamDbl$new(id = "subsample", lower = 0, upper = 1, tags = "train"),
ParamFct$new(
id = "sampling_frequency",
levels = c("PerTree", "PerTreeLevel"),
default = "PerTreeLevel", tags = "train"),
ParamFct$new(
id = "sampling_unit", levels = c("Object", "Group"),
default = "Object", tags = "train"),
ParamDbl$new(id = "mvs_reg", lower = 0, upper = Inf, tags = "train"),
ParamDbl$new(
id = "random_strength", lower = 0, upper = Inf,
default = 1, tags = "train"),
# use_best_model missing
# best_model_min_trees missing
ParamInt$new(
id = "depth", lower = 1L, upper = 16L,
default = 6L, tags = "train"),
ParamFct$new(
id = "grow_policy",
levels = c("SymmetricTree", "Depthwise", "Lossguide"),
default = "SymmetricTree", tags = "train"),
ParamInt$new(
id = "min_data_in_leaf", lower = 1L, upper = Inf,
default = 1L, tags = "train"),
ParamInt$new(
id = "max_leaves", lower = 1L, upper = Inf,
default = 31L, tags = "train"),
# ignored_features missing
# one_hot_max_size missing
ParamLgl$new(id = "has_time", default = FALSE, tags = "train"),
ParamDbl$new(
id = "rsm", lower = 0.001, upper = 1,
default = 1, tags = "train"),
ParamFct$new(
id = "nan_mode", levels = c("Min", "Max"),
default = "Min", tags = "train"), # do not allow "Forbidden"
ParamInt$new(
id = "fold_permutation_block",
lower = 1L, upper = 256L, tags = "train"),
ParamFct$new(
id = "leaf_estimation_method",
levels = c("Newton", "Gradient", "Exact"), tags = "train"),
ParamInt$new(
id = "leaf_estimation_iterations",
lower = 1L, upper = Inf, tags = "train"),
ParamFct$new(
id = "leaf_estimation_backtracking",
levels = c("No", "AnyImprovement", "Armijo"),
default = "AnyImprovement", tags = "train"),
# name missing
ParamDbl$new(
id = "fold_len_multiplier",
lower = 1.001, upper = Inf, default = 2, tags = "train"),
ParamLgl$new(
id = "approx_on_full_history",
default = TRUE, tags = "train"),
ParamUty$new(id = "class_weights", tags = "train"),
ParamFct$new(
id = "auto_class_weights",
levels = c("None", "Balanced", "SqrtBalanced"),
default = "None", tags = "train"),
ParamFct$new(
id = "boosting_type",
levels = c("Ordered", "Plain"), tags = "train"),
ParamLgl$new("boost_from_average", tags = "train"),
ParamLgl$new(id = "langevin", default = FALSE, tags = "train"),
ParamDbl$new(
id = "diffusion_temperature", lower = 0, upper = Inf,
default = 10000, tags = "train"),
# allow_const_label missing
ParamFct$new(
id = "score_function",
levels = c("Cosine", "L2", "NewtonCosine", "NewtonL2"),
default = "Cosine", tags = "train"),
# cat_features missing
ParamUty$new(
id = "monotone_constraints", tags = "train",
custom_check = checkmate::check_string),
ParamUty$new(
id = "feature_weights", tags = "train",
custom_check = checkmate::check_string),
ParamUty$new(
id = "first_feature_use_penalties", tags = "train",
custom_check = checkmate::check_string),
ParamDbl$new(
id = "penalties_coefficient", lower = 0, upper = Inf,
default = 1, tags = "train"),
ParamUty$new(
id = "per_object_feature_penalties", tags = "train",
custom_check = checkmate::check_string),
ParamDbl$new(id = "model_shrink_rate", tags = "train"),
ParamFct$new(
id = "model_shrink_mode",
levels = c("Constant", "Decreasing"), tags = "train"),
# Overfitting detection settings missing
# Quantization settings
ParamDbl$new(id = "target_border", tags = "train"),
ParamInt$new(
id = "border_count",
lower = 1L, upper = 65535L, tags = "train"),
ParamFct$new(
id = "feature_border_type",
levels = c(
"Median", "Uniform", "UniformAndQuantiles", "MaxLogSum",
"MinEntropy", "GreedyLogSum"),
default = "GreedyLogSum", tags = "train"),
ParamUty$new(
id = "per_float_feature_quantization", tags = "train",
custom_check = checkmate::check_string),
# Multiclassification settings
ParamInt$new(
id = "classes_count",
lower = 1L, upper = Inf, tags = "train"),
# Performance Settings
ParamInt$new(
id = "thread_count",
lower = -1L, upper = Inf, default = 1L,
tags = c("train", "predict", "importance")),
# Processing units settings
ParamFct$new(
id = "task_type", levels = c("CPU", "GPU"),
default = "CPU", tags = "train"),
ParamUty$new(id = "devices", tags = "train"),
# Output settings
ParamFct$new(
id = "logging_level",
levels = c("Silent", "Verbose", "Info", "Debug"),
default = "Silent", tags = "train"),
ParamInt$new(
id = "metric_period", lower = 1L, upper = Inf,
default = 1L, tags = "train"),
# verbose missing
ParamUty$new(
id = "train_dir", default = "catboost_info",
tags = "train", custom_check = checkmate::check_string),
ParamDbl$new(
id = "model_size_reg", lower = 0, upper = 1,
default = 0.5, tags = "train"),
ParamLgl$new(
id = "allow_writing_files",
default = FALSE, tags = "train"),
ParamLgl$new(id = "save_snapshot", default = FALSE, tags = "train"),
ParamUty$new(
id = "snapshot_file", tags = "train",
custom_check = checkmate::check_string),
ParamInt$new(
id = "snapshot_interval", lower = 1L, upper = Inf,
default = 600L, tags = "train"),
# CTR settings
ParamUty$new(
id = "simple_ctr", tags = "train",
custom_check = checkmate::check_string),
ParamUty$new(
id = "combinations_ctr", tags = "train",
custom_check = checkmate::check_string),
ParamInt$new(
id = "ctr_target_border_count",
lower = 1L, upper = 255L, tags = "train"),
ParamFct$new(
id = "counter_calc_method",
levels = c("SkipTest", "Full"), default = "Full", tags = "train"),
ParamInt$new(
id = "max_ctr_complexity",
lower = 1L, upper = Inf, tags = "train"),
ParamInt$new(
id = "ctr_leaf_count_limit",
lower = 1L, upper = Inf, tags = "train"),
ParamLgl$new(
id = "store_all_simple_ctr",
default = FALSE, tags = "train"),
ParamFct$new(
id = "final_ctr_computation_mode",
levels = c("Default", "Skip"), default = "Default", tags = "train"),
# catboost.predict
# https://catboost.ai/docs/concepts/r-reference_catboost-predict.html
ParamLgl$new(id = "verbose", default = FALSE, tags = "predict"),
ParamInt$new(
id = "ntree_start", lower = 0L, upper = Inf,
default = 0L, tags = "predict"),
ParamInt$new(
id = "ntree_end", lower = 0L, upper = Inf,
default = 0L, tags = "predict")
)
)
ps$add_dep(
id = "mvs_reg", on = "bootstrap_type",
cond = CondEqual$new("MVS"))
ps$add_dep(
id = "min_data_in_leaf", on = "grow_policy",
cond = CondAnyOf$new(c("Depthwise", "Lossguide")))
ps$add_dep(
id = "diffusion_temperature", on = "langevin",
cond = CondEqual$new(TRUE))
ps$values$loss_function_twoclass = "Logloss"
ps$values$loss_function_multiclass = "MultiClass"
ps$values$logging_level = "Silent"
ps$values$thread_count = 1L
ps$values$allow_writing_files = FALSE
ps$values$save_snapshot = FALSE
super$initialize(
id = "classif.catboost",
packages = "catboost",
feature_types = c("logical", "integer", "numeric", "factor", "ordered"),
predict_types = c("response", "prob"),
param_set = ps,
properties = c(
"missings", "weights", "importance", "twoclass", "multiclass"), # FIXME: parallel
man = "mlr3learners.catboost::mlr_learners_classif.catboost"
)
},
#' @description
#' The importance scores are calculated using
#' [`catboost.get_feature_importance`][catboost::catboost.get_feature_importance],
#' setting `type = "FeatureImportance"`, returned for 'all'.
#' @return Named `numeric()`.
importance = function() {
# https://catboost.ai/docs/concepts/r-reference_catboost-get_feature_importance.html
imp = mlr3misc::invoke(catboost::catboost.get_feature_importance,
model = self$model,
type = "FeatureImportance",
thread_count = self$param_set$values$thread_count)
names(imp) = self$state$train_task$feature_names
sort(imp, decreasing = TRUE)
}
),
private = list(
.train = function(task) {
# integer/logical features must be converted to numerics explicitly
data = task$data(cols = task$feature_names)
to_numerics = task$feature_types$id[task$feature_types$type %in%
c("integer", "logical")]
if (length(to_numerics)) {
data[, (to_numerics) := lapply(.SD, as.numeric), .SDcols = to_numerics]
}
# target is encoded as integer values from 0
# if binary, the positive class is 1
is_binary = (length(task$class_names) == 2L)
label = if (is_binary) {
ifelse(task$data(cols = task$target_names)[[1L]] == task$positive,
yes = 1L,
no = 0L)
} else {
as.integer(task$data(cols = task$target_names)[[1L]]) - 1L
}
# data must be a dataframe
learn_pool = mlr3misc::invoke(catboost::catboost.load_pool,
data = setDF(data),
label = label,
weight = task$weights$weight,
thread_count = self$param_set$values$thread_count)
# set loss_function correctly
pars = self$param_set$get_values(tags = "train")
pars$loss_function = if (is_binary) {
pars$loss_function_twoclass
} else {
pars$loss_function_multiclass
}
pars$loss_function_twoclass = NULL
pars$loss_function_multiclass = NULL
mlr3misc::invoke(catboost::catboost.train,
learn_pool = learn_pool,
test_pool = NULL,
params = pars)
},
.predict = function(task) {
# integer/logical features must be converted to numerics explicitly
data = task$data(cols = task$feature_names)
to_numerics = task$feature_types$id[task$feature_types$type %in%
c("integer", "logical")]
if (length(to_numerics)) {
data[, (to_numerics) := lapply(.SD, as.numeric), .SDcols = to_numerics]
}
# target was encoded as integer values based on the train_task
# to later revert this, again use the train_task
is_binary = (length(self$state$train_task$class_names) == 2L)
# data must be a dataframe
pool = mlr3misc::invoke(catboost::catboost.load_pool,
data = setDF(data),
thread_count = self$param_set$values$thread_count)
prediction_type = if (self$predict_type == "response") {
"Class"
} else {
"Probability"
}
preds = mlr3misc::invoke(catboost::catboost.predict,
model = self$model,
pool = pool,
prediction_type = prediction_type,
.args = self$param_set$get_values(tags = "predict"))
if (self$predict_type == "response") {
response = if (is_binary) {
ifelse(preds == 1L,
yes = self$state$train_task$positive,
no = setdiff(
self$state$train_task$class_names,
self$state$train_task$positive))
} else {
self$state$train_task$class_names[preds + 1L]
}
PredictionClassif$new(task = task, response = response)
} else {
if (is_binary && is.null(dim(preds))) {
preds = matrix(c(preds, 1 - preds), ncol = 2L, nrow = length(preds))
}
colnames(preds) = self$state$train_task$class_names
PredictionClassif$new(task = task, prob = preds)
}
}
)
)
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.