validate_search_data¶

`yohou.utils.validation.validate_search_data(y, X_actual)` ¶

Validate input data for hyperparameter search (GridSearchCV, RandomizedSearchCV).

Performs comprehensive validation of time series data for cross-validation: - Checks that y is not None - Validates time column presence, dtype, nulls, and sorting - Validates panel data internal consistency - Validates panel data group matching between y and X_actual - Validates consistent time intervals across DataFrames

This function is designed for SearchCV contexts where we validate data without modifying forecaster state (unlike validate_forecaster_data).

Parameters¶

Name	Type	Description	Default
`y`	`DataFrame`	Target time series with "time" column.	required
`X_actual`	`DataFrame or None`	Exogenous feature time series with "time" column, or None.	required

Returns¶

Type	Description
`str`	The common time interval shared by all provided DataFrames (e.g., "1d", "1mo").

Raises¶

Type	Description
`ValueError`	If y is None, time columns are invalid, panel data is inconsistent, or intervals don't match across DataFrames.

Examples¶

>>> import polars as pl
>>> from datetime import datetime
>>> time_index = pl.datetime_range(
...     start=datetime(2020, 1, 1), end=datetime(2020, 1, 5), interval="1d", eager=True
... )
>>> y = pl.DataFrame({"time": time_index, "sales": [100, 110, 120, 130, 140]})
>>> X_actual = pl.DataFrame({"time": time_index, "holiday": [0, 0, 1, 0, 0]})
>>> interval = validate_search_data(y, X_actual)
>>> interval
'1d'

Source Code¶

View on GitHub

Show/Hide sourcedef validate_search_data(y: pl.DataFrame, X_actual: pl.DataFrame | None) -> str:
    """Validate input data for hyperparameter search (GridSearchCV, RandomizedSearchCV).

    Performs comprehensive validation of time series data for cross-validation:
    - Checks that y is not None
    - Validates time column presence, dtype, nulls, and sorting
    - Validates panel data internal consistency
    - Validates panel data group matching between y and X_actual
    - Validates consistent time intervals across DataFrames

    This function is designed for SearchCV contexts where we validate data
    without modifying forecaster state (unlike validate_forecaster_data).

    Parameters
    ----------
    y : pl.DataFrame
        Target time series with "time" column.

    X_actual : pl.DataFrame or None
        Exogenous feature time series with "time" column, or None.

    Returns
    -------
    str
        The common time interval shared by all provided DataFrames (e.g., "1d", "1mo").

    Raises
    ------
    ValueError
        If y is None, time columns are invalid, panel data is inconsistent,
        or intervals don't match across DataFrames.

    Examples
    --------
    >>> import polars as pl
    >>> from datetime import datetime
    >>> time_index = pl.datetime_range(
    ...     start=datetime(2020, 1, 1), end=datetime(2020, 1, 5), interval="1d", eager=True
    ... )
    >>> y = pl.DataFrame({"time": time_index, "sales": [100, 110, 120, 130, 140]})
    >>> X_actual = pl.DataFrame({"time": time_index, "holiday": [0, 0, 1, 0, 0]})
    >>> interval = validate_search_data(y, X_actual)
    >>> interval
    '1d'

    See Also
    --------
    - [`validate_forecaster_data`][yohou.utils.validate_data.validate_forecaster_data] : Data validation with forecaster state management
    - [`check_inputs`][yohou.utils.validation.check_inputs] : Validates consistent time intervals
    - [`check_time_column`][yohou.utils.validation.check_time_column] : Validates time column properties

    """
    if y is None:
        raise ValueError("`y` cannot be None")

    # Validate time columns
    check_time_column(y, "y")
    if X_actual is not None:
        check_time_column(X_actual, "X_actual")

    # Validate panel data internal consistency
    check_panel_internal_consistency(y, "y")
    if X_actual is not None:
        check_panel_internal_consistency(X_actual, "X_actual")
        # Validate panel data groups match
        check_panel_groups_match(y, X_actual)

    # Validate consistent time intervals and return the interval
    return check_inputs(y, X_actual)