Sample weights

Provide a column to pass through to the underlying models as sample weights

Data setup

import numpy as np
from mlforecast.utils import generate_daily_series

series = generate_daily_series(2)
series['weight'] = np.random.default_rng(seed=0).random(series.shape[0])
series.head(2)

	unique_id	ds	y	weight
0	id_0	2000-01-01	0.357595	0.636962
1	id_0	2000-01-02	1.301382	0.269787

Creating forecast object

import lightgbm as lgb
from sklearn.linear_model import LinearRegression

from mlforecast import MLForecast

fcst = MLForecast(
    models={
        'lr': LinearRegression(),
        'lgbm': lgb.LGBMRegressor(verbosity=-1),
    },
    freq='D',
    lags=[1],
    date_features=['dayofweek'],
)

Forecasting

You can provide the weight_col argument to MLForecast.fit to indicate which column should be used as the sample weights.

fcst.fit(series, weight_col='weight').predict(1)

	unique_id	ds	lr	lgbm
0	id_0	2000-08-10	3.336019	5.283677
1	id_1	2000-04-07	3.300786	4.230655

Cross validation

You can provide the weight_col argument to MLForecast.cross_validation to indicate which column should be used as the sample weights.

fcst.cross_validation(series, n_windows=2, h=1, weight_col='weight')

	unique_id	ds	cutoff	y	lr	lgbm
0	id_0	2000-08-08	2000-08-07	3.436325	2.770717	3.242790
1	id_1	2000-04-05	2000-04-04	2.430276	2.687932	2.075247
2	id_0	2000-08-09	2000-08-08	4.136771	3.095140	4.239010
3	id_1	2000-04-06	2000-04-05	3.363522	3.016661	3.436962