def __init__(
self,
ml_task,
X_train,
y_train,
sample_weight,
X_validation,
y_validation,
sample_weight_validation,
eval_metric,
cat_features_indices,
n_jobs,
random_state,
):
self.ml_task = ml_task
self.X_train = X_train
self.y_train = y_train
self.sample_weight = sample_weight
self.X_validation = X_validation
self.y_validation = y_validation
self.eval_metric = eval_metric
self.cat_features = cat_features_indices
self.eval_set = Pool(
data=X_validation,
label=y_validation,
cat_features=self.cat_features,
weight=sample_weight_validation,
)
self.n_jobs = n_jobs
self.rounds = 1000
self.learning_rate = 0.0125
self.early_stopping_rounds = 50
self.seed = random_state
self.objective = catboost_objective(ml_task, self.eval_metric.name)
self.eval_metric_name = catboost_eval_metric(ml_task, self.eval_metric.name)
self.custom_eval_metric = None
if self.eval_metric_name == "spearman":
self.custom_eval_metric = CatBoostEvalMetricSpearman()
elif self.eval_metric_name == "pearson":
self.custom_eval_metric = CatBoostEvalMetricPearson()
elif self.eval_metric_name == "average_precision":
self.custom_eval_metric = CatBoostEvalMetricAveragePrecision()
elif self.eval_metric_name == "mse":
self.custom_eval_metric = CatBoostEvalMetricMSE()
elif self.eval_metric_name == "user_defined_metric":
self.custom_eval_metric = CatBoostEvalMetricUserDefined()
def __init__(self, params):
super(CatBoostAlgorithm, self).__init__(params)
self.library_version = catboost.__version__
self.snapshot_file_path = "training_snapshot"
self.explain_level = params.get("explain_level", 0)
self.rounds = additional.get("max_rounds", 10000)
self.max_iters = 1
self.early_stopping_rounds = additional.get("early_stopping_rounds",
50)
Algo = CatBoostClassifier
loss_function = "Logloss"
if self.params["ml_task"] == BINARY_CLASSIFICATION:
loss_function = self.params.get("loss_function", "Logloss")
elif self.params["ml_task"] == MULTICLASS_CLASSIFICATION:
loss_function = self.params.get("loss_function", "MultiClass")
elif self.params["ml_task"] == REGRESSION:
loss_function = self.params.get("loss_function", "RMSE")
Algo = CatBoostRegressor
cat_params = {
"iterations": self.params.get("num_boost_round", self.rounds),
"learning_rate": self.params.get("learning_rate", 0.1),
"depth": self.params.get("depth", 3),
"rsm": self.params.get("rsm", 1.0),
"l2_leaf_reg": self.params.get("l2_leaf_reg", 3.0),
"random_strength": self.params.get("random_strength", 1.0),
"loss_function": loss_function,
"eval_metric": self.params.get("eval_metric", loss_function),
# "custom_metric": self.params.get("eval_metric", loss_function),
"thread_count": self.params.get("n_jobs", -1),
"verbose": False,
"allow_writing_files": False,
"random_seed": self.params.get("seed", 1),
}
for extra_param in [
"min_data_in_leaf",
"bootstrap_type",
"bagging_temperature",
"subsample",
"border_count",
]:
if extra_param in self.params:
cat_params[extra_param] = self.params[extra_param]
self.log_metric_name = cat_params["eval_metric"]
if cat_params["eval_metric"] == "spearman":
cat_params["eval_metric"] = CatBoostEvalMetricSpearman()
self.log_metric_name = "CatBoostEvalMetricSpearman"
elif cat_params["eval_metric"] == "pearson":
cat_params["eval_metric"] = CatBoostEvalMetricPearson()
self.log_metric_name = "CatBoostEvalMetricPearson"
elif cat_params["eval_metric"] == "average_precision":
cat_params["eval_metric"] = CatBoostEvalMetricAveragePrecision()
self.log_metric_name = "CatBoostEvalMetricAveragePrecision"
elif cat_params["eval_metric"] == "mse":
cat_params["eval_metric"] = CatBoostEvalMetricMSE()
self.log_metric_name = "CatBoostEvalMetricMSE"
self.model = Algo(**cat_params)
self.cat_features = None
self.best_ntree_limit = 0
logger.debug("CatBoostAlgorithm.__init__")