OLSModel

Ordinary Least Squares linear regression using scikit-learn’s LinearRegression.

Basic Usage

from epftoolbox2.models import OLSModel

model = OLSModel(
    predictors=[
        "load_actual",
        "is_monday_d+{horizon}",
        "is_tuesday_d+{horizon}",
        "is_wednesday_d+{horizon}",
        "is_thursday_d+{horizon}",
        "is_friday_d+{horizon}",
        "is_saturday_d+{horizon}",
        "is_sunday_d+{horizon}",
        "is_holiday_d+{horizon}",
        "daylight_hours_d+{horizon}",
    ],
    training_window=365,
    name="OLS",
)

Parameters

Parameter	Type	Default	Description
`predictors`	List	Required	Predictor specifications (strings, `{horizon}` templates, or callables)
`training_window`	int	`365`	Days of training data
`name`	str	`"Model"`	Display name for reports

Callable predictors accept (horizon) or (horizon, hour) — the two-parameter form enables features that differ by hour of day:

# horizon-only
lambda horizon: f"price_lag_d-{horizon-7}"

# horizon + hour of day
lambda horizon, hour: f"load_h{hour:02d}_d+{horizon}"

Example

from epftoolbox2.pipelines import ModelPipeline
from epftoolbox2.models import OLSModel
from epftoolbox2.evaluators import MAEEvaluator
from epftoolbox2.exporters import TerminalExporter

predictors = [
    "load_actual",
    "is_monday_d+{horizon}",
    "is_tuesday_d+{horizon}",
    "is_wednesday_d+{horizon}",
    "is_thursday_d+{horizon}",
    "is_friday_d+{horizon}",
    "is_saturday_d+{horizon}",
    "is_sunday_d+{horizon}",
    "is_holiday_d+{horizon}",
    "daylight_hours_d+{horizon}",
    *[f"load_actual_d-{i}" for i in [1, 7]],
]

pipeline = (
    ModelPipeline()
    .add_model(OLSModel(predictors=predictors, name="OLS"))
    .add_evaluator(MAEEvaluator())
    .add_exporter(TerminalExporter())
)

report = pipeline.run(data=df, test_start="2024-02-01", test_end="2024-03-01")