Sklearn models

from functools import partial

from datasetsforecast.m4 import M4, M4Info
from sklearn.linear_model import Lasso, Ridge
from utilsforecast.feature_engineering import pipeline, trend, fourier
from utilsforecast.plotting import plot_series

from statsforecast import StatsForecast
from statsforecast.models import SklearnModel
from statsforecast.utils import ConformalIntervals

group = 'Hourly'
season_length = M4Info[group].seasonality
horizon = M4Info[group].horizon
data, *_ = M4.load('data', group)
data['ds'] = data['ds'].astype('int64')
valid = data.groupby('unique_id').tail(horizon).copy()
train = data.drop(valid.index)
train.head()

train_features, valid_features = pipeline(
    train,
    features=[
        trend,        
        partial(fourier, season_length=season_length, k=10),  # 10 fourier terms
    ],
    freq=1,
    h=horizon,
)
train_features.head()

sf = StatsForecast(
    models=[
        SklearnModel(Lasso()),
        SklearnModel(Ridge()),
    ],
    freq=1,
)
preds = sf.forecast(
    df=train_features,
    h=horizon,
    X_df=valid_features,
    prediction_intervals=ConformalIntervals(n_windows=4, h=horizon),
    level=[95],
)
plot_series(train, preds, level=[95], palette='tab20b', max_ids=4)