Yandex
CatBoost
 
Help
Overview of CatBoost
Installation
Python package
R package
Command-line version
Applying models
Objectives and metrics
Model analysis
Data format description
Parameter tuning
Speeding up the training
Data visualization
Educational materials
Development and contributions
Reference
Contacts
Overview of CatBoost
Installation
Python package installation
pip install
conda install
Build from source on Linux and macOS
Build from source on Windows
Build a wheel package
Additional packages for data visualization support
Test CatBoost
R package installation
Install the released version
conda install
Build from source
Install from a local copy on Linux and macOS
Install from a local copy on Windows
Command-line version binary
Download
Build the binary from a local copy on Linux and macOS
Build the binary from a local copy on Windows
Build the binary with make on Linux (CPU only)
Build the binary with MPI support from a local copy (GPU only)
CatBoost Viewer installation
Python package
Quick start
Pool
num_row
num_col
get_features
get_label
get_cat_feature_indices
get_weight
get_baseline
set_pairs
set_feature_names
set_baseline
set_weight
set_group_id
set_group_weight
set_subgroup_id
set_pairs_weight
slice
Attributes
Pool initialization
FeaturesData
get_object_count
get_num_feature_count
get_cat_feature_count
get_feature_count
get_feature_names
CatBoost
fit
predict
staged_predict
eval_metrics
get_feature_importance
get_object_importance
load_model
save_model
shrink
get_param
get_params
set_params
copy
get_evals_result
get_test_eval
is_fitted
get_metadata
get_best_score
get_best_iteration
Attributes
CatBoostClassifier
fit
predict
predict_proba
staged_predict
staged_predict_proba
eval_metrics
get_feature_importance
get_object_importance
load_model
save_model
shrink
get_param
get_params
set_params
score
copy
get_evals_result
get_test_eval
is_fitted
get_metadata
get_best_score
get_best_iteration
Attributes
CatBoostRegressor
fit
predict
staged_predict
eval_metrics
get_feature_importance
get_object_importance
load_model
save_model
shrink
get_param
get_params
set_params
score
copy
get_evals_result
get_test_eval
is_fitted
get_metadata
get_best_score
get_best_iteration
Attributes
train
cv
datasets
titanic
amazon
adult
epsilon
MetricVisualizer
start
sum_models
utils
create_cd
eval_metric
get_gpu_device_count
get_roc_curve
get_fpr_curve
get_fnr_curve
select_threshold
Training parameters
Pool initialization
Missing values processing
Usage examples
R package
Quick start
catboost.load_pool
catboost.save_pool
catboost.train
catboost.load_model
catboost.save_model
catboost.predict
catboost.shrink
catboost.staged_predict
catboost.get_feature_importance
catboost.get_object_importance
catboost.get_model_params
Training parameters
Attributes
Missing values processing
Usage examples
Command-line version
Train a model
Cross-validation
Apply a model
Calculate metrics
Calculate feature importance
Calculate object importance
Metadata manipulation
Sum models
Missing values processing
Usage examples
Applying models
Evaluator based on a C/C++ library
Standalone C++ evaluator
Using models exported as Python code
Apply in Java
CatBoostModel
loadModel
getPredictionDimension
getTreeCount
getUsedCategoricFeatureCount
getUsedNumericFeatureCount
predict
close
CatBoostPredictions
copyRowMajorPredictions
copyObjectPredictions
get
getObjectCount
getPredictionDimension
Using models exported as C++ code
Exporting the model to Apple CoreML
Using models exported to ONNX-ML format
Objectives and metrics
Regression
Classification
Multiclassification
Ranking
Model analysis
Feature importance
Object importance
Data format description
Input data
Column descriptions
Dataset description
Pair description
Custom quantization borders and missing value modes
Group weights
Output data
Model values
Feature analysis
Feature importance
Feature interaction strength
ShapValues
Objects strength
Metrics and time information
Profiler information
Metric
Time information
stdout
Custom quantization borders and missing value modes
ROC curve points
Parameter tuning
Speeding up the training
Data visualization
Jupyter Notebook
CatBoost Viewer
TensorBoard
Educational materials
Tutorials
Reference papers
Videos
Development and contributions
Reference
How training is performed
Preliminary calculation of splits
Transforming categorical features to numerical features
Choosing the tree structure
Unbiased boosting
Binarization
Overfitting detector
Recovering training after an interruption
Contacts

С and C++ API

The following methods for applying a trained model are provided:
  • Evaluator based on a C/C++ library
  • Standalone C++ evaluator

The C API and all constant methods of the TFullModel class are thread-safe.