Source code for evalml.data_checks.ts_parameters_data_check

"""Data check that checks whether the time series parameters are compatible with the data size."""
from evalml.data_checks import DataCheck, DataCheckError, DataCheckMessageCode
from evalml.utils.gen_utils import (
    are_ts_parameters_valid_for_split,
    contains_all_ts_parameters,
)


[docs]class TimeSeriesParametersDataCheck(DataCheck): """Checks whether the time series parameters are compatible with data splitting. If `gap + max_delay + forecast_horizon > X.shape[0] // (n_splits + 1)` then the feature engineering window is larger than the smallest split. This will cause the pipeline to create features from data that does not exist, which will cause errors. Args: problem_configuration (dict): Dict containing problem_configuration parameters. n_splits (int): Number of time series splits. """ def __init__(self, problem_configuration, n_splits): is_valid, msg = contains_all_ts_parameters(problem_configuration) if not is_valid: raise ValueError(msg) self.gap = problem_configuration["gap"] self.forecast_horizon = problem_configuration["forecast_horizon"] self.max_delay = problem_configuration["max_delay"] self.n_splits = n_splits
[docs] def validate(self, X, y=None): """Check if the time series parameters are compatible with data splitting. Args: X (pd.DataFrame, np.ndarray): Features. y (pd.Series, np.ndarray): Ignored. Defaults to None. Returns: dict: dict with a DataCheckError if parameters are too big for the split sizes. Examples: >>> import pandas as pd The time series parameters have to be compatible with the data passed. If the window size (gap + max_delay + forecast_horizon) is greater than or equal to the split size, then an error will be raised. >>> X = pd.DataFrame({ ... "dates": pd.date_range("1/1/21", periods=100), ... "first": [i for i in range(100)], ... }) >>> y = pd.Series([i for i in range(100)]) ... >>> problem_config = {"gap": 7, "max_delay": 2, "forecast_horizon": 12, "time_index": "dates"} >>> target_leakage_check = TimeSeriesParametersDataCheck(problem_configuration=problem_config, n_splits=4) >>> assert target_leakage_check.validate(X, y) == [ ... { ... "message": "Since the data has 100 observations and n_splits=4, the smallest " ... "split would have 20 observations. Since 21 (gap + max_delay + forecast_horizon)" ... " >= 20, then at least one of the splits would be empty by the time it reaches " ... "the pipeline. Please use a smaller number of splits, reduce one or more these " ... "parameters, or collect more data.", ... "data_check_name": "TimeSeriesParametersDataCheck", ... "level": "error", ... "code": "TIMESERIES_PARAMETERS_NOT_COMPATIBLE_WITH_SPLIT", ... "details": { ... "columns": None, ... "rows": None, ... "max_window_size": 21, ... "min_split_size": 20, ... "n_obs": 100, ... "n_splits": 4 ... }, ... "action_options": [] ... } ... ] """ messages = [] validation = are_ts_parameters_valid_for_split( gap=self.gap, max_delay=self.max_delay, forecast_horizon=self.forecast_horizon, n_splits=self.n_splits, n_obs=X.shape[0], ) if not validation.is_valid: messages.append( DataCheckError( message=validation.msg, data_check_name=self.name, message_code=DataCheckMessageCode.TIMESERIES_PARAMETERS_NOT_COMPATIBLE_WITH_SPLIT, details={ "max_window_size": validation.max_window_size, "min_split_size": validation.smallest_split_size, "n_obs": validation.n_obs, "n_splits": validation.n_splits, }, ).to_dict() ) return messages