import numpy as np
import pandas as pd
from evalml.model_family import ModelFamily
from evalml.pipelines.components.estimators import Estimator
from evalml.problem_types import ProblemTypes
from evalml.utils import infer_feature_types, pad_with_nans
[docs]class TimeSeriesBaselineEstimator(Estimator):
"""Time series estimator that predicts using the naive forecasting approach.
This is useful as a simple baseline estimator for time series problems.
Arguments:
gap (int): Gap between prediction date and target date and must be a positive integer. If gap is 0, target date will be shifted ahead by 1 time period. Defaults to 1.
random_seed (int): Seed for the random number generator. Defaults to 0.
"""
name = "Time Series Baseline Estimator"
hyperparameter_ranges = {}
"""{}"""
model_family = ModelFamily.BASELINE
"""ModelFamily.BASELINE"""
supported_problem_types = [
ProblemTypes.TIME_SERIES_REGRESSION,
ProblemTypes.TIME_SERIES_BINARY,
ProblemTypes.TIME_SERIES_MULTICLASS,
]
"""[
ProblemTypes.TIME_SERIES_REGRESSION,
ProblemTypes.TIME_SERIES_BINARY,
ProblemTypes.TIME_SERIES_MULTICLASS,
]"""
predict_uses_y = True
def __init__(self, gap=1, random_seed=0, **kwargs):
self._prediction_value = None
self._num_features = None
self.gap = gap
if gap < 0:
raise ValueError(
f"gap value must be a positive integer. {gap} was provided."
)
parameters = {"gap": gap}
parameters.update(kwargs)
super().__init__(
parameters=parameters, component_obj=None, random_seed=random_seed
)
[docs] def fit(self, X, y=None):
if X is None:
X = pd.DataFrame()
X = infer_feature_types(X)
self._num_features = X.shape[1]
return self
[docs] def predict(self, X, y=None):
if y is None:
raise ValueError(
"Cannot predict Time Series Baseline Estimator if y is None"
)
y = infer_feature_types(y)
if self.gap == 0:
y = y.shift(periods=1)
return infer_feature_types(y)
[docs] def predict_proba(self, X, y=None):
if y is None:
raise ValueError(
"Cannot predict Time Series Baseline Estimator if y is None"
)
y = infer_feature_types(y)
preds = self.predict(X, y).dropna(axis=0, how="any").astype("int")
proba_arr = np.zeros((len(preds), y.max() + 1))
proba_arr[np.arange(len(preds)), preds] = 1
padded = pad_with_nans(pd.DataFrame(proba_arr), len(y) - len(preds))
return infer_feature_types(padded)
@property
def feature_importance(self):
"""Returns importance associated with each feature.
Since baseline estimators do not use input features to calculate predictions, returns an array of zeroes.
Returns:
np.ndarray (float): an array of zeroes
"""
return np.zeros(self._num_features)