forecaster.wrappers.lgbm

forecaster.wrappers.lgbm

Recursive forecaster wrapper using LightGBM.

Classes

Name	Description
ForecasterRecursiveLGBM	ForecasterRecursive specialization using LightGBM.

ForecasterRecursiveLGBM

forecaster.wrappers.lgbm.ForecasterRecursiveLGBM(iteration, lags=12, **kwargs)

ForecasterRecursive specialization using LightGBM.

Examples

import warnings
import numpy as np
import pandas as pd
from lightgbm import LGBMRegressor

from spotforecast2_safe.exceptions import IgnoredArgumentWarning
from spotforecast2_safe.forecaster.recursive import ForecasterRecursive
from spotforecast2_safe.forecaster.wrappers import ForecasterRecursiveLGBM

warnings.simplefilter("ignore", IgnoredArgumentWarning)

rng = np.random.default_rng(0)
n = 50
idx = pd.date_range("2023-01-01", periods=n, freq="h")
vals = 100 + 10 * np.sin(np.linspace(0, 4 * np.pi, n)) + rng.normal(0, 1, n)
y = pd.Series(vals, index=idx)

# Use a fast, small estimator for the example
model = ForecasterRecursiveLGBM(iteration=0, lags=3)
model.forecaster = ForecasterRecursive(
    estimator=LGBMRegressor(
        n_estimators=10, n_jobs=1, verbose=-1, random_state=0,
        deterministic=True, force_col_wise=True,
    ),
    lags=3,
)
model.fit(y=y)
pred = model.forecaster.predict(steps=2)

assert model.name == "lgbm"
assert model.forecaster.is_fitted
assert len(pred) == 2
assert pred.dtype == float
print(f"model.name: {model.name}")
print(f"is_fitted: {model.forecaster.is_fitted}")
print(f"prediction steps: {len(pred)}")

model.name: lgbm
is_fitted: True
prediction steps: 2