catboost – CatBoost gradient-boosted trees (optional dependency).#

Back to family axis | Back to L4 | Browse all options

Operational op under axis family, sub-layer L4_A_model_selection, layer l4. Standalone callable: mf.functions.catboost_fit.

Function signature#

mf.functions.catboost_fit(
    X: np.ndarray | pd.DataFrame,
    y: np.ndarray | pd.Series,
) -> CatBoostFitResult

Parameters#

name

type

default

constraint

description

X

`np.ndarray

pd.DataFrame`

—

—

y

`np.ndarray

pd.Series`

—

—

Returns#

CatBoostFitResult — frozen dataclass with fit results.

Attribute

Type

Description

.feature_importances_

np.ndarray

Feature importances from CatBoost (percentage-based), shape (n_features,).

.n_estimators_used

int

Number of boosting iterations (= n_estimators parameter).

.predict(X)

np.ndarray

Predictions for new data X, guaranteed 1-D via .ravel().

.summary()

str

Human-readable table of fit results including top-3 feature importances.

Behavior#

Requires pip install macroforecast[catboost]. Ordered boosting + native categorical handling.

When to use

Categorical-heavy panels; ordered-boosting research.

In recipe context#

Set params.family = "catboost" in the relevant layer to activate this op within a recipe:

# Layer L4 recipe fragment
params:
  family: catboost

References#

  • macroforecast design Part 2, L4: ‘forecasting model is the layer where every authoring iteration ends – pick family, tune, repeat.’

  • Prokhorenkova et al. (2018) ‘CatBoost: unbiased boosting with categorical features’, NeurIPS.