ForecasterAutoregCustom¶
ForecasterAutoregCustom (ForecasterBase)
¶
This class turns any regressor compatible with the scikit-learn API into a
recursive (multi-step) forecaster with a custom function to create predictors.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
regressor |
regressor or pipeline compatible with the scikit-learn API |
An instance of a regressor or pipeline compatible with the scikit-learn API. |
required |
fun_predictors |
<built-in function callable> |
Function that takes a numpy ndarray as a window of values as input and |
required |
window_size |
int |
Size of the window needed by |
required |
transformer_y |
transformer (preprocessor) compatible with the scikit-learn |
preprocessing API, default |
None |
transformer_exog |
transformer (preprocessor) compatible with the scikit-learn |
preprocessing API, default |
None |
Attributes:
| Name | Type | Description |
|---|---|---|
regressor |
regressor compatible with the scikit-learn API |
An instance of a regressor compatible with the scikit-learn API. |
create_predictors |
Callable |
Function that takes a numpy ndarray as a window of values as input and |
source_code_create_predictors |
str |
Source code of the custom function used to create the predictors. |
window_size |
int |
Size of the window needed by |
transformer_y |
transformer (preprocessor) compatible with the scikit-learn |
preprocessing API, default |
transformer_exog |
transformer (preprocessor) compatible with the scikit-learn |
preprocessing API, default |
last_window |
pandas Series |
Last window the forecaster has seen during trained. It stores the
values needed to predict the next |
window_size |
int |
Size of the window needed by |
fitted |
Bool |
Tag to identify if the regressor has been fitted (trained). |
index_type |
type |
Type of index of the input used in training. |
index_freq |
str |
Frequency of Index of the input used in training. |
training_range |
pandas Index |
First and last values of index of the data used during training. |
included_exog |
bool |
If the forecaster has been trained using exogenous variable/s. |
exog_type |
type |
Type of exogenous variable/s used in training. |
exog_col_names |
list |
Names of columns of |
X_train_col_names |
list |
Names of columns of the matrix created internally for training. |
in_sample_residuals |
pandas Series |
Residuals of the model when predicting training data. Only stored up to 1000 values. |
out_sample_residuals |
pandas Series |
Residuals of the model when predicting non training data. Only stored
up to 1000 values. Use |
creation_date |
str |
Date of creation. |
fit_date |
str |
Date of last fit. |
skforcast_version |
str |
Version of skforecast library used to create the forecaster. |
python_version |
str |
Version of python used to create the forecaster. New in version 0.5.0 |
Source code in skforecast/ForecasterAutoregCustom/ForecasterAutoregCustom.py
class ForecasterAutoregCustom(ForecasterBase):
'''
This class turns any regressor compatible with the scikit-learn API into a
recursive (multi-step) forecaster with a custom function to create predictors.
Parameters
----------
regressor : regressor or pipeline compatible with the scikit-learn API
An instance of a regressor or pipeline compatible with the scikit-learn API.
fun_predictors : Callable
Function that takes a numpy ndarray as a window of values as input and
returns a numpy ndarray with the predictors associated with that window.
window_size : int
Size of the window needed by `fun_predictors` to create the predictors.
transformer_y : transformer (preprocessor) compatible with the scikit-learn
preprocessing API, default `None`
An instance of a transformer (preprocessor) compatible with the scikit-learn
preprocessing API with methods: fit, transform, fit_transform and inverse_transform.
ColumnTransformers are not allowed since they do not have inverse_transform method.
The transformation is applied to `y` before training the forecaster.
**New in version 0.5.0**
transformer_exog : transformer (preprocessor) compatible with the scikit-learn
preprocessing API, default `None`
An instance of a transformer (preprocessor) compatible with the scikit-learn
preprocessing API. The transformation is applied to `exog` before training the
forecaster. `inverse_transform` is not available when using ColumnTransformers.
**New in version 0.5.0**
Attributes
----------
regressor : regressor compatible with the scikit-learn API
An instance of a regressor compatible with the scikit-learn API.
create_predictors : Callable
Function that takes a numpy ndarray as a window of values as input and
returns a numpy ndarray with the predictors associated with that window.
source_code_create_predictors : str
Source code of the custom function used to create the predictors.
window_size : int
Size of the window needed by `fun_predictors` to create the predictors.
transformer_y : transformer (preprocessor) compatible with the scikit-learn
preprocessing API, default `None`
An instance of a transformer (preprocessor) compatible with the scikit-learn
preprocessing API with methods: fit, transform, fit_transform and inverse_transform.
ColumnTransformers are not allowed since they do not have inverse_transform method.
The transformation is applied to `y` before training the forecaster.
**New in version 0.5.0**
transformer_exog : transformer (preprocessor) compatible with the scikit-learn
preprocessing API, default `None`
An instance of a transformer (preprocessor) compatible with the scikit-learn
preprocessing API. The transformation is applied to `exog` before training the
forecaster. `inverse_transform` is not available when using ColumnTransformers.
**New in version 0.5.0**
last_window : pandas Series
Last window the forecaster has seen during trained. It stores the
values needed to predict the next `step` right after the training data.
window_size : int
Size of the window needed by `fun_predictors` to create the predictors.
fitted : Bool
Tag to identify if the regressor has been fitted (trained).
index_type : type
Type of index of the input used in training.
index_freq : str
Frequency of Index of the input used in training.
training_range : pandas Index
First and last values of index of the data used during training.
included_exog : bool
If the forecaster has been trained using exogenous variable/s.
exog_type : type
Type of exogenous variable/s used in training.
exog_col_names : list
Names of columns of `exog` if `exog` used in training was a pandas
DataFrame.
X_train_col_names : list
Names of columns of the matrix created internally for training.
in_sample_residuals : pandas Series
Residuals of the model when predicting training data. Only stored up to
1000 values.
out_sample_residuals : pandas Series
Residuals of the model when predicting non training data. Only stored
up to 1000 values. Use `set_out_sample_residuals` to set values.
creation_date : str
Date of creation.
fit_date : str
Date of last fit.
skforcast_version : str
Version of skforecast library used to create the forecaster.
python_version : str
Version of python used to create the forecaster.
**New in version 0.5.0**
'''
def __init__(
self,
regressor,
fun_predictors: callable,
window_size: int,
transformer_y = None,
transformer_exog = None,
) -> None:
self.regressor = regressor
self.create_predictors = fun_predictors
self.source_code_create_predictors = None
self.window_size = window_size
self.transformer_y = transformer_y
self.transformer_exog = transformer_exog
self.index_type = None
self.index_freq = None
self.training_range = None
self.last_window = None
self.included_exog = False
self.exog_type = None
self.exog_col_names = None
self.X_train_col_names = None
self.in_sample_residuals = None
self.out_sample_residuals = None
self.fitted = False
self.creation_date = pd.Timestamp.today().strftime('%Y-%m-%d %H:%M:%S')
self.fit_date = None
self.skforcast_version = skforecast.__version__
self.python_version = sys.version.split(" ")[0]
if not isinstance(window_size, int):
raise Exception(
f'`window_size` must be int, got {type(window_size)}'
)
if not callable(fun_predictors):
raise Exception(
f'`fun_predictors` must be callable, got {type(fun_predictors)}.'
)
self.source_code_create_predictors = getsource(fun_predictors)
def __repr__(
self
) -> str:
"""
Information displayed when a ForecasterAutoregCustom object is printed.
"""
if isinstance(self.regressor, sklearn.pipeline.Pipeline):
name_pipe_steps = tuple(name + "__" for name in self.regressor.named_steps.keys())
params = {key : value for key, value in self.regressor.get_params().items() \
if key.startswith(name_pipe_steps)}
else:
params = self.regressor.get_params()
info = (
f"{'=' * len(str(type(self)).split('.')[1])} \n"
f"{str(type(self)).split('.')[1]} \n"
f"{'=' * len(str(type(self)).split('.')[1])} \n"
f"Regressor: {self.regressor} \n"
f"Predictors created with function: {self.create_predictors.__name__} \n"
f"Window size: {self.window_size} \n"
f"Transformer for y: {self.transformer_y} \n"
f"Transformer for exog: {self.transformer_exog} \n"
f"Included exogenous: {self.included_exog} \n"
f"Type of exogenous variable: {self.exog_type} \n"
f"Exogenous variables names: {self.exog_col_names} \n"
f"Training range: {self.training_range.to_list() if self.fitted else None} \n"
f"Training index type: {str(self.index_type).split('.')[-1][:-2] if self.fitted else None} \n"
f"Training index frequency: {self.index_freq if self.fitted else None} \n"
f"Regressor parameters: {params} \n"
f"Creation date: {self.creation_date} \n"
f"Last fit date: {self.fit_date} \n"
f"Skforecast version: {self.skforcast_version} \n"
f"Python version: {self.python_version} \n"
)
return info
def create_train_X_y(
self,
y: pd.Series,
exog: Optional[Union[pd.Series, pd.DataFrame]]=None
) -> Tuple[pd.DataFrame, pd.Series]:
"""
Create training matrices from univariate time series.
Parameters
----------
y : pandas Series
Training time series.
exog : pandas Series, pandas DataFrame, default `None`
Exogenous variable/s included as predictor/s. Must have the same
number of observations as `y` and their indexes must be aligned.
Returns
-------
X_train : pandas DataFrame
Pandas DataFrame with the training values (predictors).
y_train : pandas Series
Values (target) of the time series related to each row of `X_train`.
"""
if len(y) < self.window_size + 1:
raise Exception(
f'`y` must have as many values as the windows_size needed by '
f'{self.create_predictors.__name__}. For this Forecaster the '
f'minimum length is {self.window_size + 1}'
)
check_y(y=y)
y = transform_series(
series = y,
transformer = self.transformer_y,
fit = True,
inverse_transform = False
)
y_values, y_index = preprocess_y(y=y)
if exog is not None:
if len(exog) != len(y):
raise Exception(
"`exog` must have same number of samples as `y`."
)
check_exog(exog=exog)
if isinstance(exog, pd.Series):
exog = transform_series(
series = exog,
transformer = self.transformer_exog,
fit = True,
inverse_transform = False
)
else:
exog = transform_dataframe(
df = exog,
transformer = self.transformer_exog,
fit = True,
inverse_transform = False
)
exog_values, exog_index = preprocess_exog(exog=exog)
if not (exog_index[:len(y_index)] == y_index).all():
raise Exception(
('Different index for `y` and `exog`. They must be equal '
'to ensure the correct alignment of values.')
)
X_train = []
y_train = []
for i in range(len(y) - self.window_size):
train_index = np.arange(i, self.window_size + i)
test_index = self.window_size + i
X_train.append(self.create_predictors(y=y_values[train_index]))
y_train.append(y_values[test_index])
X_train = np.vstack(X_train)
y_train = np.array(y_train)
X_train_col_names = [f"custom_predictor_{i}" for i in range(X_train.shape[1])]
if np.isnan(X_train).any():
raise Exception(
f"`create_predictors()` is returning `NaN` values."
)
if exog is not None:
col_names_exog = exog.columns if isinstance(exog, pd.DataFrame) else [exog.name]
X_train_col_names.extend(col_names_exog)
# The first `self.window_size` positions have to be removed from exog
# since they are not in X_train.
X_train = np.column_stack((X_train, exog_values[self.window_size:, ]))
X_train = pd.DataFrame(
data = X_train,
columns = X_train_col_names,
index = y_index[self.window_size: ]
)
self.X_train_col_names = X_train_col_names
y_train = pd.Series(
data = y_train,
index = y_index[self.window_size: ],
name = 'y'
)
return X_train, y_train
def fit(
self,
y: pd.Series,
exog: Optional[Union[np.ndarray, pd.Series, pd.DataFrame]]=None
) -> None:
"""
Training Forecaster.
Parameters
----------
y : pandas Series
Training time series.
exog : pandas Series, pandas DataFrame, default `None`
Exogenous variable/s included as predictor/s. Must have the same
number of observations as `y` and their indexes must be aligned so
that y[i] is regressed on exog[i].
Returns
-------
None
"""
# Reset values in case the forecaster has already been fitted.
self.index_type = None
self.index_freq = None
self.last_window = None
self.included_exog = False
self.exog_type = None
self.exog_col_names = None
self.in_sample_residuals = None
self.X_train_col_names = None
self.fitted = False
self.training_range = None
if exog is not None:
self.included_exog = True
self.exog_type = type(exog)
self.exog_col_names = \
exog.columns.to_list() if isinstance(exog, pd.DataFrame) else exog.name
X_train, y_train = self.create_train_X_y(y=y, exog=exog)
if not str(type(self.regressor)) == "<class 'xgboost.sklearn.XGBRegressor'>":
self.regressor.fit(X=X_train, y=y_train)
else:
self.regressor.fit(X=X_train.to_numpy(), y=y_train.to_numpy())
self.fitted = True
self.fit_date = pd.Timestamp.today().strftime('%Y-%m-%d %H:%M:%S')
self.training_range = preprocess_y(y=y)[1][[0, -1]]
self.index_type = type(X_train.index)
if isinstance(X_train.index, pd.DatetimeIndex):
self.index_freq = X_train.index.freqstr
else:
self.index_freq = X_train.index.step
if not str(type(self.regressor)) == "<class 'xgboost.sklearn.XGBRegressor'>":
residuals = y_train - self.regressor.predict(X_train)
else:
residuals = y_train - self.regressor.predict(X_train.to_numpy())
residuals = pd.Series(
data = residuals,
index = y_train.index,
name = 'in_sample_residuals'
)
if len(residuals) > 1000:
# Only up to 1000 residuals are stored
residuals = residuals.sample(n=1000, random_state=123, replace=False)
self.in_sample_residuals = residuals
# The last time window of training data is stored so that predictors in
# the first iteration of `predict()` can be calculated.
self.last_window = y.iloc[-self.window_size:].copy()
def _recursive_predict(
self,
steps: int,
last_window: np.ndarray,
exog: np.ndarray
) -> np.array:
"""
Predict n steps ahead. It is an iterative process in which, each prediction,
is used as a predictor for the next step.
Parameters
----------
steps : int
Number of future steps predicted.
last_window : numpy ndarray
Values of the series used to create the predictors (lags) need in the
first iteration of prediction (t + 1).
exog : numpy ndarray, pandas DataFrame
Exogenous variable/s included as predictor/s.
Returns
-------
predictions : numpy ndarray
Predicted values.
"""
predictions = np.full(shape=steps, fill_value=np.nan)
for i in range(steps):
X = self.create_predictors(y=last_window).reshape(1, -1)
if np.isnan(X).any():
raise Exception(
f"`create_predictors()` is returning `NaN` values."
)
if exog is not None:
X = np.column_stack((X, exog[i, ].reshape(1, -1)))
with warnings.catch_warnings():
# Suppress scikitlearn warning: "X does not have valid feature names,
# but NoOpTransformer was fitted with feature names".
warnings.simplefilter("ignore")
prediction = self.regressor.predict(X)
predictions[i] = prediction.ravel()[0]
# Update `last_window` values. The first position is discarded and
# the new prediction is added at the end.
last_window = np.append(last_window[1:], prediction)
return predictions
def predict(
self,
steps: int,
last_window: Optional[pd.Series]=None,
exog: Optional[Union[pd.Series, pd.DataFrame]]=None
) -> pd.Series:
"""
Predict n steps ahead. It is an recursive process in which, each prediction,
is used as a predictor for the next step.
Parameters
----------
steps : int
Number of future steps predicted.
last_window : pandas Series, default `None`
Values of the series used to create the predictors (lags) need in the
first iteration of prediction (t + 1).
If `last_window = None`, the values stored in` self.last_window` are
used to calculate the initial predictors, and the predictions start
right after training data.
exog : pandas Series, pandas DataFrame, default `None`
Exogenous variable/s included as predictor/s.
Returns
-------
predictions : pandas Series
Predicted values.
"""
check_predict_input(
forecaster_type = type(self),
steps = steps,
fitted = self.fitted,
included_exog = self.included_exog,
index_type = self.index_type,
index_freq = self.index_freq,
window_size = self.window_size,
last_window = last_window,
exog = exog,
exog_type = self.exog_type,
exog_col_names = self.exog_col_names,
interval = None,
max_steps = None,
level = None,
series_levels = None
)
if exog is not None:
if isinstance(exog, pd.DataFrame):
exog = transform_dataframe(
df = exog,
transformer = self.transformer_exog,
fit = False,
inverse_transform = False
)
else:
exog = transform_series(
series = exog,
transformer = self.transformer_exog,
fit = False,
inverse_transform = False
)
exog_values, _ = preprocess_exog(
exog = exog.iloc[:steps, ]
)
else:
exog_values = None
if last_window is None:
last_window = self.last_window.copy()
last_window = transform_series(
series = last_window,
transformer = self.transformer_y,
fit = False,
inverse_transform = False
)
last_window_values, last_window_index = preprocess_last_window(
last_window = last_window
)
predictions = self._recursive_predict(
steps = steps,
last_window = copy(last_window_values),
exog = copy(exog_values)
)
predictions = pd.Series(
data = predictions,
index = expand_index(
index = last_window_index,
steps = steps
),
name = 'pred'
)
predictions = transform_series(
series = predictions,
transformer = self.transformer_y,
fit = False,
inverse_transform = True
)
return predictions
def _estimate_boot_interval(
self,
steps: int,
last_window: Optional[np.ndarray]=None,
exog: Optional[np.ndarray]=None,
interval: list=[5, 95],
n_boot: int=500,
random_state: int=123,
in_sample_residuals: bool=True
) -> np.ndarray:
"""
Iterative process in which, each prediction, is used as a predictor
for the next step and bootstrapping is used to estimate prediction
intervals. This method only returns prediction intervals.
See predict_intervals() to calculate both, predictions and intervals.
Parameters
----------
steps : int
Number of future steps predicted.
last_window : 1d numpy ndarray shape (, max_lag), default `None`
Values of the series used to create the predictors (lags) needed in the
first iteration of prediction (t + 1).
If `last_window = None`, the values stored in` self.last_window` are
used to calculate the initial predictors, and the predictions start
right after training data.
exog : numpy ndarray, default `None`
Exogenous variable/s included as predictor/s.
n_boot : int, default `500`
Number of bootstrapping iterations used to estimate prediction
intervals.
random_state : int
Sets a seed to the random generator, so that boot intervals are always
deterministic.
interval : list, default `[5, 95]`
Confidence of the prediction interval estimated. Sequence of
percentiles to compute, which must be between 0 and 100 inclusive.
For example, interval of 95% should be as `interval = [2.5, 97.5]`.
in_sample_residuals : bool, default `True`
If `True`, residuals from the training data are used as proxy of
prediction error to create prediction intervals. If `False`, out of
sample residuals are used. In the latter case, the user should have
calculated and stored the residuals within the forecaster (see
`set_out_sample_residuals()`).
Returns
-------
prediction_interval : numpy ndarray, shape (steps, 2)
Interval estimated for each prediction by bootstrapping:
- lower_bound: lower bound of the interval.
- upper_bound: upper bound interval of the interval.
Notes
-----
More information about prediction intervals in forecasting:
https://otexts.com/fpp2/prediction-intervals.html
Forecasting: Principles and Practice (2nd ed) Rob J Hyndman and
George Athanasopoulos.
"""
if last_window is None:
last_window = self.last_window.values
boot_predictions = np.full(
shape = (steps, n_boot),
fill_value = np.nan,
dtype = float
)
rng = np.random.default_rng(seed=random_state)
seeds = rng.integers(low=0, high=10000, size=n_boot)
for i in range(n_boot):
# In each bootstraping iteration the initial last_window and exog
# need to be restored.
last_window_boot = last_window.copy()
if exog is not None:
exog_boot = exog.copy()
else:
exog_boot = None
if in_sample_residuals:
residuals = self.in_sample_residuals
else:
residuals = self.out_sample_residuals
rng = np.random.default_rng(seed=seeds[i])
sample_residuals = rng.choice(
a = residuals,
size = steps,
replace = True
)
for step in range(steps):
prediction = self._recursive_predict(
steps = 1,
last_window = last_window_boot,
exog = exog_boot
)
prediction_with_residual = prediction + sample_residuals[step]
boot_predictions[step, i] = prediction_with_residual
last_window_boot = np.append(
last_window_boot[1:],
prediction_with_residual
)
if exog is not None:
exog_boot = exog_boot[1:]
prediction_interval = np.percentile(boot_predictions, q=interval, axis=1)
prediction_interval = prediction_interval.transpose()
return prediction_interval
def predict_interval(
self,
steps: int,
last_window: Optional[pd.Series]=None,
exog: Optional[Union[pd.Series, pd.DataFrame]]=None,
interval: list=[5, 95],
n_boot: int=500,
random_state: int=123,
in_sample_residuals: bool=True
) -> pd.DataFrame:
"""
Iterative process in which, each prediction, is used as a predictor
for the next step and bootstrapping is used to estimate prediction
intervals. Both, predictions and intervals, are returned.
Parameters
----------
steps : int
Number of future steps predicted.
last_window : pandas Series, default `None`
Values of the series used to create the predictors (lags) needed in the
first iteration of prediction (t + 1).
If `last_window = None`, the values stored in` self.last_window` are
used to calculate the initial predictors, and the predictions start
right after training data.
exog : pandas Series, pandas DataFrame, default `None`
Exogenous variable/s included as predictor/s.
interval : list, default `[5, 95]`
Confidence of the prediction interval estimated. Sequence of
percentiles to compute, which must be between 0 and 100 inclusive.
For example, interval of 95% should be as `interval = [2.5, 97.5]`.
n_boot : int, default `500`
Number of bootstrapping iterations used to estimate prediction
intervals.
random_state : int, default 123
Sets a seed to the random generator, so that boot intervals are always
deterministic.
in_sample_residuals : bool, default `True`
If `True`, residuals from the training data are used as proxy of
prediction error to create prediction intervals. If `False`, out of
sample residuals are used. In the latter case, the user should have
calculated and stored the residuals within the forecaster (see
`set_out_sample_residuals()`).
Returns
-------
predictions : pandas DataFrame
Values predicted by the forecaster and their estimated interval:
- pred: predictions.
- lower_bound: lower bound of the interval.
- upper_bound: upper bound interval of the interval.
Notes
-----
More information about prediction intervals in forecasting:
https://otexts.com/fpp2/prediction-intervals.html
Forecasting: Principles and Practice (2nd ed) Rob J Hyndman and
George Athanasopoulos.
"""
check_predict_input(
forecaster_type = type(self),
steps = steps,
fitted = self.fitted,
included_exog = self.included_exog,
index_type = self.index_type,
index_freq = self.index_freq,
window_size = self.window_size,
last_window = last_window,
exog = exog,
exog_type = self.exog_type,
exog_col_names = self.exog_col_names,
interval = interval,
max_steps = None,
level = None,
series_levels = None
)
if exog is not None:
if isinstance(exog, pd.DataFrame):
exog = transform_dataframe(
df = exog,
transformer = self.transformer_exog,
fit = False,
inverse_transform = False
)
else:
exog = transform_series(
series = exog,
transformer = self.transformer_exog,
fit = False,
inverse_transform = False
)
exog_values, _ = preprocess_exog(
exog = exog.iloc[:steps, ]
)
else:
exog_values = None
if last_window is None:
last_window = self.last_window.copy()
last_window = transform_series(
series = last_window,
transformer = self.transformer_y,
fit = False,
inverse_transform = False
)
last_window_values, last_window_index = preprocess_last_window(
last_window = last_window
)
# Since during predict() `last_window` and `exog` are modified, the
# originals are stored to be used later
last_window_values_original = last_window_values.copy()
if exog is not None:
exog_values_original = exog_values.copy()
else:
exog_values_original = None
predictions = self._recursive_predict(
steps = steps,
last_window = last_window_values,
exog = exog_values
)
predictions_interval = self._estimate_boot_interval(
steps = steps,
last_window = copy(last_window_values_original),
exog = copy(exog_values_original),
interval = interval,
n_boot = n_boot,
random_state = random_state,
in_sample_residuals = in_sample_residuals
)
predictions = np.column_stack((predictions, predictions_interval))
predictions = pd.DataFrame(
data = predictions,
index = expand_index(
index = last_window_index,
steps = steps
),
columns = ['pred', 'lower_bound', 'upper_bound']
)
if self.transformer_y:
for col in predictions.columns:
predictions[col] = self.transformer_y.inverse_transform(predictions[[col]])
return predictions
def set_params(
self,
**params: dict
) -> None:
"""
Set new values to the parameters of the scikit learn model stored in the
ForecasterAutoregCustom.
Parameters
----------
params : dict
Parameters values.
Returns
-------
self
"""
self.regressor = clone(self.regressor)
self.regressor.set_params(**params)
def set_out_sample_residuals(
self,
residuals: pd.Series,
append: bool=True,
transform: bool=True
)-> None:
"""
Set new values to the attribute `out_sample_residuals`. Out of sample
residuals are meant to be calculated using observations that did not
participate in the training process.
Parameters
----------
residuals : pd.Series
Values of residuals. If len(residuals) > 1000, only a random sample
of 1000 values are stored.
append : bool, default `True`
If `True`, new residuals are added to the once already stored in the
attribute `out_sample_residuals`. Once the limit of 1000 values is
reached, no more values are appended. If False, `out_sample_residuals`
is overwritten with the new residuals.
transform : bool, default `True`
If `True`, new residuals are transformed using self.transformer_y.
Returns
-------
self
"""
if not isinstance(residuals, pd.Series):
raise Exception(
f"`residuals` argument must be `pd.Series`. Got {type(residuals)}"
)
if not transform and self.transformer_y is not None:
warnings.warn(
f'''
Argument `transform` is set to `False` but forecaster was trained
using a transformer {self.transformer_y}. Ensure that new residuals
are already transformed or set `transform=True`.
'''
)
if transform and self.transformer_y is not None:
warnings.warn(
f'''
Residuals will be transformed using the same transformer used
when training the forecaster ({self.transformer_y}). Ensure that
new residuals are in the same scale as the original time series.
'''
)
residuals = transform_series(
series = residuals,
transformer = self.transformer_y,
fit = False,
inverse_transform = False
)
if len(residuals) > 1000:
rng = np.random.default_rng(seed=123)
residuals = rng.choice(a=residuals, size=1000, replace=False)
residuals = pd.Series(residuals)
if append and self.out_sample_residuals is not None:
free_space = max(0, 1000 - len(self.out_sample_residuals))
if len(residuals) < free_space:
residuals = np.hstack((
self.out_sample_residuals,
residuals
))
else:
residuals = np.hstack((
self.out_sample_residuals,
residuals[:free_space]
))
self.out_sample_residuals = pd.Series(residuals)
def get_feature_importance(
self
) -> pd.DataFrame:
"""
Return feature importance of the regressor stored in the
forecaster. Only valid when regressor stores internally the feature
importance in the attribute `feature_importances_` or `coef_`.
Parameters
----------
self
Returns
-------
feature_importance : pandas DataFrame
Feature importance associated with each predictor.
"""
if self.fitted == False:
raise sklearn.exceptions.NotFittedError(
"This forecaster is not fitted yet. Call `fit` with appropriate "
"arguments before using `get_feature_importance()`."
)
if isinstance(self.regressor, sklearn.pipeline.Pipeline):
estimator = self.regressor[-1]
else:
estimator = self.regressor
try:
feature_importance = pd.DataFrame({
'feature': self.X_train_col_names,
'importance' : estimator.feature_importances_
})
except:
try:
feature_importance = pd.DataFrame({
'feature': self.X_train_col_names,
'importance' : estimator.coef_
})
except:
warnings.warn(
f"Impossible to access feature importance for regressor of type {type(estimator)}. "
f"This method is only valid when the regressor stores internally "
f"the feature importance in the attribute `feature_importances_` "
f"or `coef_`."
)
feature_importance = None
return feature_importance
create_train_X_y(self, y, exog=None)
¶
Create training matrices from univariate time series.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
y |
Series |
Training time series. |
required |
exog |
Union[pandas.core.series.Series, pandas.core.frame.DataFrame] |
Exogenous variable/s included as predictor/s. Must have the same
number of observations as |
None |
Returns:
| Type | Description |
|---|---|
Tuple[pandas.core.frame.DataFrame, pandas.core.series.Series] |
Pandas DataFrame with the training values (predictors). |
Source code in skforecast/ForecasterAutoregCustom/ForecasterAutoregCustom.py
def create_train_X_y(
self,
y: pd.Series,
exog: Optional[Union[pd.Series, pd.DataFrame]]=None
) -> Tuple[pd.DataFrame, pd.Series]:
"""
Create training matrices from univariate time series.
Parameters
----------
y : pandas Series
Training time series.
exog : pandas Series, pandas DataFrame, default `None`
Exogenous variable/s included as predictor/s. Must have the same
number of observations as `y` and their indexes must be aligned.
Returns
-------
X_train : pandas DataFrame
Pandas DataFrame with the training values (predictors).
y_train : pandas Series
Values (target) of the time series related to each row of `X_train`.
"""
if len(y) < self.window_size + 1:
raise Exception(
f'`y` must have as many values as the windows_size needed by '
f'{self.create_predictors.__name__}. For this Forecaster the '
f'minimum length is {self.window_size + 1}'
)
check_y(y=y)
y = transform_series(
series = y,
transformer = self.transformer_y,
fit = True,
inverse_transform = False
)
y_values, y_index = preprocess_y(y=y)
if exog is not None:
if len(exog) != len(y):
raise Exception(
"`exog` must have same number of samples as `y`."
)
check_exog(exog=exog)
if isinstance(exog, pd.Series):
exog = transform_series(
series = exog,
transformer = self.transformer_exog,
fit = True,
inverse_transform = False
)
else:
exog = transform_dataframe(
df = exog,
transformer = self.transformer_exog,
fit = True,
inverse_transform = False
)
exog_values, exog_index = preprocess_exog(exog=exog)
if not (exog_index[:len(y_index)] == y_index).all():
raise Exception(
('Different index for `y` and `exog`. They must be equal '
'to ensure the correct alignment of values.')
)
X_train = []
y_train = []
for i in range(len(y) - self.window_size):
train_index = np.arange(i, self.window_size + i)
test_index = self.window_size + i
X_train.append(self.create_predictors(y=y_values[train_index]))
y_train.append(y_values[test_index])
X_train = np.vstack(X_train)
y_train = np.array(y_train)
X_train_col_names = [f"custom_predictor_{i}" for i in range(X_train.shape[1])]
if np.isnan(X_train).any():
raise Exception(
f"`create_predictors()` is returning `NaN` values."
)
if exog is not None:
col_names_exog = exog.columns if isinstance(exog, pd.DataFrame) else [exog.name]
X_train_col_names.extend(col_names_exog)
# The first `self.window_size` positions have to be removed from exog
# since they are not in X_train.
X_train = np.column_stack((X_train, exog_values[self.window_size:, ]))
X_train = pd.DataFrame(
data = X_train,
columns = X_train_col_names,
index = y_index[self.window_size: ]
)
self.X_train_col_names = X_train_col_names
y_train = pd.Series(
data = y_train,
index = y_index[self.window_size: ],
name = 'y'
)
return X_train, y_train
fit(self, y, exog=None)
¶
Training Forecaster.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
y |
Series |
Training time series. |
required |
exog |
Union[numpy.ndarray, pandas.core.series.Series, pandas.core.frame.DataFrame] |
Exogenous variable/s included as predictor/s. Must have the same
number of observations as |
None |
Source code in skforecast/ForecasterAutoregCustom/ForecasterAutoregCustom.py
def fit(
self,
y: pd.Series,
exog: Optional[Union[np.ndarray, pd.Series, pd.DataFrame]]=None
) -> None:
"""
Training Forecaster.
Parameters
----------
y : pandas Series
Training time series.
exog : pandas Series, pandas DataFrame, default `None`
Exogenous variable/s included as predictor/s. Must have the same
number of observations as `y` and their indexes must be aligned so
that y[i] is regressed on exog[i].
Returns
-------
None
"""
# Reset values in case the forecaster has already been fitted.
self.index_type = None
self.index_freq = None
self.last_window = None
self.included_exog = False
self.exog_type = None
self.exog_col_names = None
self.in_sample_residuals = None
self.X_train_col_names = None
self.fitted = False
self.training_range = None
if exog is not None:
self.included_exog = True
self.exog_type = type(exog)
self.exog_col_names = \
exog.columns.to_list() if isinstance(exog, pd.DataFrame) else exog.name
X_train, y_train = self.create_train_X_y(y=y, exog=exog)
if not str(type(self.regressor)) == "<class 'xgboost.sklearn.XGBRegressor'>":
self.regressor.fit(X=X_train, y=y_train)
else:
self.regressor.fit(X=X_train.to_numpy(), y=y_train.to_numpy())
self.fitted = True
self.fit_date = pd.Timestamp.today().strftime('%Y-%m-%d %H:%M:%S')
self.training_range = preprocess_y(y=y)[1][[0, -1]]
self.index_type = type(X_train.index)
if isinstance(X_train.index, pd.DatetimeIndex):
self.index_freq = X_train.index.freqstr
else:
self.index_freq = X_train.index.step
if not str(type(self.regressor)) == "<class 'xgboost.sklearn.XGBRegressor'>":
residuals = y_train - self.regressor.predict(X_train)
else:
residuals = y_train - self.regressor.predict(X_train.to_numpy())
residuals = pd.Series(
data = residuals,
index = y_train.index,
name = 'in_sample_residuals'
)
if len(residuals) > 1000:
# Only up to 1000 residuals are stored
residuals = residuals.sample(n=1000, random_state=123, replace=False)
self.in_sample_residuals = residuals
# The last time window of training data is stored so that predictors in
# the first iteration of `predict()` can be calculated.
self.last_window = y.iloc[-self.window_size:].copy()
get_feature_importance(self)
¶
Return feature importance of the regressor stored in the
forecaster. Only valid when regressor stores internally the feature
importance in the attribute feature_importances_ or coef_.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
self |
None |
required |
Returns:
| Type | Description |
|---|---|
DataFrame |
Feature importance associated with each predictor. |
Source code in skforecast/ForecasterAutoregCustom/ForecasterAutoregCustom.py
def get_feature_importance(
self
) -> pd.DataFrame:
"""
Return feature importance of the regressor stored in the
forecaster. Only valid when regressor stores internally the feature
importance in the attribute `feature_importances_` or `coef_`.
Parameters
----------
self
Returns
-------
feature_importance : pandas DataFrame
Feature importance associated with each predictor.
"""
if self.fitted == False:
raise sklearn.exceptions.NotFittedError(
"This forecaster is not fitted yet. Call `fit` with appropriate "
"arguments before using `get_feature_importance()`."
)
if isinstance(self.regressor, sklearn.pipeline.Pipeline):
estimator = self.regressor[-1]
else:
estimator = self.regressor
try:
feature_importance = pd.DataFrame({
'feature': self.X_train_col_names,
'importance' : estimator.feature_importances_
})
except:
try:
feature_importance = pd.DataFrame({
'feature': self.X_train_col_names,
'importance' : estimator.coef_
})
except:
warnings.warn(
f"Impossible to access feature importance for regressor of type {type(estimator)}. "
f"This method is only valid when the regressor stores internally "
f"the feature importance in the attribute `feature_importances_` "
f"or `coef_`."
)
feature_importance = None
return feature_importance
predict(self, steps, last_window=None, exog=None)
¶
Predict n steps ahead. It is an recursive process in which, each prediction,
is used as a predictor for the next step.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
steps |
int |
Number of future steps predicted. |
required |
last_window |
Optional[pandas.core.series.Series] |
Values of the series used to create the predictors (lags) need in the first iteration of prediction (t + 1). If |
None |
exog |
Union[pandas.core.series.Series, pandas.core.frame.DataFrame] |
Exogenous variable/s included as predictor/s. |
None |
Returns:
| Type | Description |
|---|---|
Series |
Predicted values. |
Source code in skforecast/ForecasterAutoregCustom/ForecasterAutoregCustom.py
def predict(
self,
steps: int,
last_window: Optional[pd.Series]=None,
exog: Optional[Union[pd.Series, pd.DataFrame]]=None
) -> pd.Series:
"""
Predict n steps ahead. It is an recursive process in which, each prediction,
is used as a predictor for the next step.
Parameters
----------
steps : int
Number of future steps predicted.
last_window : pandas Series, default `None`
Values of the series used to create the predictors (lags) need in the
first iteration of prediction (t + 1).
If `last_window = None`, the values stored in` self.last_window` are
used to calculate the initial predictors, and the predictions start
right after training data.
exog : pandas Series, pandas DataFrame, default `None`
Exogenous variable/s included as predictor/s.
Returns
-------
predictions : pandas Series
Predicted values.
"""
check_predict_input(
forecaster_type = type(self),
steps = steps,
fitted = self.fitted,
included_exog = self.included_exog,
index_type = self.index_type,
index_freq = self.index_freq,
window_size = self.window_size,
last_window = last_window,
exog = exog,
exog_type = self.exog_type,
exog_col_names = self.exog_col_names,
interval = None,
max_steps = None,
level = None,
series_levels = None
)
if exog is not None:
if isinstance(exog, pd.DataFrame):
exog = transform_dataframe(
df = exog,
transformer = self.transformer_exog,
fit = False,
inverse_transform = False
)
else:
exog = transform_series(
series = exog,
transformer = self.transformer_exog,
fit = False,
inverse_transform = False
)
exog_values, _ = preprocess_exog(
exog = exog.iloc[:steps, ]
)
else:
exog_values = None
if last_window is None:
last_window = self.last_window.copy()
last_window = transform_series(
series = last_window,
transformer = self.transformer_y,
fit = False,
inverse_transform = False
)
last_window_values, last_window_index = preprocess_last_window(
last_window = last_window
)
predictions = self._recursive_predict(
steps = steps,
last_window = copy(last_window_values),
exog = copy(exog_values)
)
predictions = pd.Series(
data = predictions,
index = expand_index(
index = last_window_index,
steps = steps
),
name = 'pred'
)
predictions = transform_series(
series = predictions,
transformer = self.transformer_y,
fit = False,
inverse_transform = True
)
return predictions
predict_interval(self, steps, last_window=None, exog=None, interval=[5, 95], n_boot=500, random_state=123, in_sample_residuals=True)
¶
Iterative process in which, each prediction, is used as a predictor
for the next step and bootstrapping is used to estimate prediction intervals. Both, predictions and intervals, are returned.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
steps |
int |
Number of future steps predicted. |
required |
last_window |
Optional[pandas.core.series.Series] |
Values of the series used to create the predictors (lags) needed in the first iteration of prediction (t + 1). If |
None |
exog |
Union[pandas.core.series.Series, pandas.core.frame.DataFrame] |
Exogenous variable/s included as predictor/s. |
None |
interval |
list |
Confidence of the prediction interval estimated. Sequence of
percentiles to compute, which must be between 0 and 100 inclusive.
For example, interval of 95% should be as |
[5, 95] |
n_boot |
int |
Number of bootstrapping iterations used to estimate prediction intervals. |
500 |
random_state |
int |
Sets a seed to the random generator, so that boot intervals are always deterministic. |
123 |
in_sample_residuals |
bool |
If |
True |
Returns:
| Type | Description |
|---|---|
DataFrame |
Values predicted by the forecaster and their estimated interval:
|
Source code in skforecast/ForecasterAutoregCustom/ForecasterAutoregCustom.py
def predict_interval(
self,
steps: int,
last_window: Optional[pd.Series]=None,
exog: Optional[Union[pd.Series, pd.DataFrame]]=None,
interval: list=[5, 95],
n_boot: int=500,
random_state: int=123,
in_sample_residuals: bool=True
) -> pd.DataFrame:
"""
Iterative process in which, each prediction, is used as a predictor
for the next step and bootstrapping is used to estimate prediction
intervals. Both, predictions and intervals, are returned.
Parameters
----------
steps : int
Number of future steps predicted.
last_window : pandas Series, default `None`
Values of the series used to create the predictors (lags) needed in the
first iteration of prediction (t + 1).
If `last_window = None`, the values stored in` self.last_window` are
used to calculate the initial predictors, and the predictions start
right after training data.
exog : pandas Series, pandas DataFrame, default `None`
Exogenous variable/s included as predictor/s.
interval : list, default `[5, 95]`
Confidence of the prediction interval estimated. Sequence of
percentiles to compute, which must be between 0 and 100 inclusive.
For example, interval of 95% should be as `interval = [2.5, 97.5]`.
n_boot : int, default `500`
Number of bootstrapping iterations used to estimate prediction
intervals.
random_state : int, default 123
Sets a seed to the random generator, so that boot intervals are always
deterministic.
in_sample_residuals : bool, default `True`
If `True`, residuals from the training data are used as proxy of
prediction error to create prediction intervals. If `False`, out of
sample residuals are used. In the latter case, the user should have
calculated and stored the residuals within the forecaster (see
`set_out_sample_residuals()`).
Returns
-------
predictions : pandas DataFrame
Values predicted by the forecaster and their estimated interval:
- pred: predictions.
- lower_bound: lower bound of the interval.
- upper_bound: upper bound interval of the interval.
Notes
-----
More information about prediction intervals in forecasting:
https://otexts.com/fpp2/prediction-intervals.html
Forecasting: Principles and Practice (2nd ed) Rob J Hyndman and
George Athanasopoulos.
"""
check_predict_input(
forecaster_type = type(self),
steps = steps,
fitted = self.fitted,
included_exog = self.included_exog,
index_type = self.index_type,
index_freq = self.index_freq,
window_size = self.window_size,
last_window = last_window,
exog = exog,
exog_type = self.exog_type,
exog_col_names = self.exog_col_names,
interval = interval,
max_steps = None,
level = None,
series_levels = None
)
if exog is not None:
if isinstance(exog, pd.DataFrame):
exog = transform_dataframe(
df = exog,
transformer = self.transformer_exog,
fit = False,
inverse_transform = False
)
else:
exog = transform_series(
series = exog,
transformer = self.transformer_exog,
fit = False,
inverse_transform = False
)
exog_values, _ = preprocess_exog(
exog = exog.iloc[:steps, ]
)
else:
exog_values = None
if last_window is None:
last_window = self.last_window.copy()
last_window = transform_series(
series = last_window,
transformer = self.transformer_y,
fit = False,
inverse_transform = False
)
last_window_values, last_window_index = preprocess_last_window(
last_window = last_window
)
# Since during predict() `last_window` and `exog` are modified, the
# originals are stored to be used later
last_window_values_original = last_window_values.copy()
if exog is not None:
exog_values_original = exog_values.copy()
else:
exog_values_original = None
predictions = self._recursive_predict(
steps = steps,
last_window = last_window_values,
exog = exog_values
)
predictions_interval = self._estimate_boot_interval(
steps = steps,
last_window = copy(last_window_values_original),
exog = copy(exog_values_original),
interval = interval,
n_boot = n_boot,
random_state = random_state,
in_sample_residuals = in_sample_residuals
)
predictions = np.column_stack((predictions, predictions_interval))
predictions = pd.DataFrame(
data = predictions,
index = expand_index(
index = last_window_index,
steps = steps
),
columns = ['pred', 'lower_bound', 'upper_bound']
)
if self.transformer_y:
for col in predictions.columns:
predictions[col] = self.transformer_y.inverse_transform(predictions[[col]])
return predictions
set_out_sample_residuals(self, residuals, append=True, transform=True)
¶
Set new values to the attribute out_sample_residuals. Out of sample
residuals are meant to be calculated using observations that did not participate in the training process.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
residuals |
Series |
Values of residuals. If len(residuals) > 1000, only a random sample of 1000 values are stored. |
required |
append |
bool |
If |
True |
transform |
bool |
If |
True |
Source code in skforecast/ForecasterAutoregCustom/ForecasterAutoregCustom.py
def set_out_sample_residuals(
self,
residuals: pd.Series,
append: bool=True,
transform: bool=True
)-> None:
"""
Set new values to the attribute `out_sample_residuals`. Out of sample
residuals are meant to be calculated using observations that did not
participate in the training process.
Parameters
----------
residuals : pd.Series
Values of residuals. If len(residuals) > 1000, only a random sample
of 1000 values are stored.
append : bool, default `True`
If `True`, new residuals are added to the once already stored in the
attribute `out_sample_residuals`. Once the limit of 1000 values is
reached, no more values are appended. If False, `out_sample_residuals`
is overwritten with the new residuals.
transform : bool, default `True`
If `True`, new residuals are transformed using self.transformer_y.
Returns
-------
self
"""
if not isinstance(residuals, pd.Series):
raise Exception(
f"`residuals` argument must be `pd.Series`. Got {type(residuals)}"
)
if not transform and self.transformer_y is not None:
warnings.warn(
f'''
Argument `transform` is set to `False` but forecaster was trained
using a transformer {self.transformer_y}. Ensure that new residuals
are already transformed or set `transform=True`.
'''
)
if transform and self.transformer_y is not None:
warnings.warn(
f'''
Residuals will be transformed using the same transformer used
when training the forecaster ({self.transformer_y}). Ensure that
new residuals are in the same scale as the original time series.
'''
)
residuals = transform_series(
series = residuals,
transformer = self.transformer_y,
fit = False,
inverse_transform = False
)
if len(residuals) > 1000:
rng = np.random.default_rng(seed=123)
residuals = rng.choice(a=residuals, size=1000, replace=False)
residuals = pd.Series(residuals)
if append and self.out_sample_residuals is not None:
free_space = max(0, 1000 - len(self.out_sample_residuals))
if len(residuals) < free_space:
residuals = np.hstack((
self.out_sample_residuals,
residuals
))
else:
residuals = np.hstack((
self.out_sample_residuals,
residuals[:free_space]
))
self.out_sample_residuals = pd.Series(residuals)
set_params(self, **params)
¶
Set new values to the parameters of the scikit learn model stored in the
ForecasterAutoregCustom.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
params |
dict |
Parameters values. |
{} |
Source code in skforecast/ForecasterAutoregCustom/ForecasterAutoregCustom.py
def set_params(
self,
**params: dict
) -> None:
"""
Set new values to the parameters of the scikit learn model stored in the
ForecasterAutoregCustom.
Parameters
----------
params : dict
Parameters values.
Returns
-------
self
"""
self.regressor = clone(self.regressor)
self.regressor.set_params(**params)