Skip to main content

CatBoost Example

Train and Save​

from catboost import CatBoostClassifier
from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split

data = load_breast_cancer()
X_train, X_test, y_train, y_test = train_test_split(
    data.data, data.target, test_size=0.2, random_state=42
)

model = CatBoostClassifier(iterations=50, depth=4, random_seed=42, verbose=0)
model.fit(X_train, y_train)

# Must export as JSON for Timber
model.save_model("catboost_model.json", format="json")

Load and Serve​

CatBoost JSON shares the .json extension with XGBoost, so use --format to disambiguate:

timber load catboost_model.json --name catboost-bc --format catboost

Timber can also auto-detect CatBoost by looking for the oblivious_trees key in the JSON.

timber serve catboost-bc

CatBoost Oblivious Trees​

CatBoost uses oblivious decision trees — symmetric trees where every node at the same depth uses the same split feature and threshold. Timber's parser expands these into the general tree representation used by the IR, converting the bottom-up split ordering to the standard top-down format.