def _predict_proba_drcif(self, X, X_p, X_d, c22, n_intervals, intervals,
dims, atts):
"""Embedded predict proba for the DrCIF classifier."""
if not self._is_fitted:
raise NotFittedError(
f"This instance of {self.__class__.__name__} has not "
f"been fitted yet; please call `fit` first.")
X = check_X(X, coerce_to_numpy=True)
n_instances, n_dims, series_length = X.shape
dists = np.zeros((n_instances, self.n_classes))
for i in range(n_instances):
r = [
X[i].reshape((1, n_dims, series_length)),
X_p[i].reshape((1, n_dims, X_p.shape[2])),
X_d[i].reshape((1, n_dims, X_d.shape[2])),
]
dists[i] = self.root.predict_proba_drcif(
r,
c22,
n_intervals,
intervals,
dims,
atts,
self.n_classes,
)
return dists
def check_is_fitted(self, method_name=None):
"""Check if `fit` has been called.
Parameters
----------
method_name : str
Name of the calling method.
Raises
------
NotFittedError
If forecaster has not been fitted yet.
"""
super(BaseGridSearch, self).check_is_fitted()
# We additionally check if the tuned forecaster has been fitted.
if method_name is not None:
if not self.refit:
raise NotFittedError(
"This %s instance was initialized "
"with refit=False. %s is "
"available only after refitting on the "
"best parameters. You can refit an forecaster "
"manually using the ``best_params_`` "
"attribute" % (type(self).__name__, method_name))
else:
self.best_forecaster_.check_is_fitted()
def predict_proba(self, X):
"""Probability estimates for each class for all cases in X.
Parameters
----------
X : The training input samples. array-like or sparse matrix of shape
= [n_test_instances,n_attributes]
Returns
-------
output : array of shape = [n_test_instances, num_classes] of
probabilities
"""
if not self._is_fitted:
raise NotFittedError(
f"This instance of {self.__class__.__name__} has not "
f"been fitted yet; please call `fit` first.")
if isinstance(X, np.ndarray) and len(X.shape) == 3 and X.shape[1] == 1:
X = np.reshape(X, (X.shape[0], -1))
elif not isinstance(X, np.ndarray) or len(X.shape) > 2:
raise ValueError(
"ContinuousIntervalTree is not a time series classifier. "
"A 2d numpy array is required.")
dists = np.zeros((X.shape[0], self.n_classes))
for i in range(X.shape[0]):
dists[i] = self.root.predict_proba(X[i], self.n_classes)
return dists
def check_is_fitted(self):
"""Check if the estimator has been fitted.
Raises
------
NotFittedError
If the estimator has not been fitted yet.
"""
if not self.is_fitted:
raise NotFittedError(
f"This instance of {self.__class__.__name__} has not "
f"been fitted yet; please call `fit` first.")
def check_is_fitted(self, method_name=None):
super(BaseGridSearch, self).check_is_fitted()
if method_name is not None:
if not self.refit:
raise NotFittedError("This %s instance was initialized "
"with refit=False. %s is "
"available only after refitting on the "
"best "
"parameters. You can refit an forecaster "
"manually using the ``best_params_`` "
"attribute" %
(type(self).__name__, method_name))
else:
self.best_forecaster_.check_is_fitted()
def _get_train_probs(self, X, y):
if not self._is_fitted:
raise NotFittedError(
f"This instance of {self.__class__.__name__} has not "
f"been fitted yet; please call `fit` first.")
if isinstance(X, np.ndarray) and len(X.shape) == 3 and X.shape[1] == 1:
X = np.reshape(X, (X.shape[0], -1))
elif isinstance(X, pd.DataFrame) and len(X.shape) == 2:
X = X.to_numpy()
elif not isinstance(X, np.ndarray) or len(X.shape) > 2:
raise ValueError("RotationForest is not a time series classifier. "
"A 2d numpy array is required.")
n_instances, n_atts = X.shape
if n_instances != self.n_instances or n_atts != self.n_atts:
raise ValueError(
"n_instances, n_atts mismatch. X should be the same as the training "
"data used in fit for generating train probabilities.")
if not self.save_transformed_data:
raise ValueError(
"Currently only works with saved transform data from fit.")
p = Parallel(n_jobs=self._n_jobs)(
delayed(self._train_probas_for_estimator)(
y,
i,
) for i in range(self._n_estimators))
y_probas, oobs = zip(*p)
results = np.sum(y_probas, axis=0)
divisors = np.zeros(n_instances)
for oob in oobs:
for inst in oob:
divisors[inst] += 1
for i in range(n_instances):
results[i] = (np.ones(self.n_classes) * (1 / self.n_classes)
if divisors[i] == 0 else results[i] /
(np.ones(self.n_classes) * divisors[i]))
return results
def check_is_fitted(estimator, msg=None):
"""Perform is_fitted validation for estimator.
Adapted from sklearn.utils.validation.check_is_fitted
Checks if the estimator is fitted by verifying the presence and
positivity of self.is_fitted_
Parameters
----------
msg : string
The default error message is, "This %(name)s instance is not fitted
yet. Call 'fit' with appropriate arguments before using this
estimator."
For custom messages if "%(name)s" is present in the message string,
it is substituted for the estimator name.
Eg. : "Estimator, %(name)s, must be fitted before sparsifying".
Returns
-------
None
Raises
------
NotFittedError
If the attributes are not found.
"""
if isclass(estimator):
raise TypeError("{} is a class, not an instance.".format(estimator))
if msg is None:
msg = (
"This %(name)s instance is not fitted yet. Call 'fit' with "
"appropriate arguments before using this estimator."
)
if not hasattr(estimator, "fit"):
raise TypeError("%s is not an estimator instance." % (estimator))
if not hasattr(estimator, "_is_fitted") or not estimator.is_fitted:
raise NotFittedError(msg % {"name": type(estimator).__name__})
def predict_proba(self, X):
"""Probability estimates for each class for all cases in X.
Parameters
----------
X : ndarray of shape = [n_instances,n_attributes]
Returns
-------
output : array of shape = [n_test_instances, num_classes] of
probabilities
"""
if not self._is_fitted:
raise NotFittedError(
f"This instance of {self.__class__.__name__} has not "
f"been fitted yet; please call `fit` first.")
if isinstance(X, np.ndarray) and len(X.shape) == 3 and X.shape[1] == 1:
X = np.reshape(X, (X.shape[0], -1))
elif isinstance(X, pd.DataFrame) and len(X.shape) == 2:
X = X.to_numpy()
elif not isinstance(X, np.ndarray) or len(X.shape) > 2:
raise ValueError("RotationForest is not a time series classifier. "
"A 2d numpy array is required.")
# replace missing values with 0 and remove useless attributes
X = np.nan_to_num(X, False, 0, 0, 0)
X = X[:, self._useful_atts]
# normalise the data.
X = (X - self._min) / self._ptp
y_probas = Parallel(n_jobs=self._n_jobs)(
delayed(self._predict_proba_for_estimator)(
X,
self.estimators_[i],
self._pcas[i],
self._groups[i],
) for i in range(self._n_estimators))
output = np.sum(
y_probas, axis=0) / (np.ones(self.n_classes) * self._n_estimators)
return output
def predict_proba_cif(self, X, c22, intervals, dims, atts):
if not self._is_fitted:
raise NotFittedError(
f"This instance of {self.__class__.__name__} has not "
f"been fitted yet; please call `fit` first.")
X = check_X(X, coerce_to_numpy=True)
n_instances, n_dims, series_length = X.shape
dists = np.zeros((n_instances, self.n_classes))
for i in range(n_instances):
dists[i] = self.root.predict_proba_cif(
X[i].reshape((1, n_dims, series_length)),
c22,
intervals,
dims,
atts,
self.n_classes,
self.class_dictionary,
)
return dists
