def weighted_sum(self, X):
""" Get class votes from the sum of rules that fires.
The rules are weighted.
parameters
----------
X: numpy.ndarray of length equal to the number of features.
Instance attributes.
Returns
-------
dict (class_value, weight)
The class distribution from the sum of the fired rules.
"""
final_votes = {}
fired_rule = False
for rule in self.rule_set:
if rule.covers_instance(X):
fired_rule = True
votes = copy.deepcopy(rule.get_class_votes(X, self))
if sum(votes.values()) != 0:
votes = normalize_values_in_dict(votes, inplace=False)
final_votes = {k: final_votes.get(k, 0) + votes.get(k, 0) for k in set(final_votes) | set(votes)}
if sum(final_votes.values()) != 0:
normalize_values_in_dict(final_votes)
return final_votes if fired_rule else self.default_rule.get_class_votes(X, self)
def get_class_votes(self, X, ht):
# dist = {}
prediction_option = ht.leaf_prediction
# MC
if prediction_option == ht._MAJORITY_CLASS:
dist = self.get_observed_class_distribution()
# NB
elif prediction_option == ht._NAIVE_BAYES:
dist = do_naive_bayes_prediction(X, self._observed_class_distribution,
self._attribute_observers)
# NBAdaptive (default)
else:
if self._mc_correct_weight > self._nb_correct_weight:
dist = self.get_observed_class_distribution()
else:
dist = do_naive_bayes_prediction(X, self._observed_class_distribution,
self._attribute_observers)
dist_sum = sum(dist.values()) # sum all values in dictionary
normalization_factor = dist_sum * self.get_error_estimation() * self.get_error_estimation()
if normalization_factor > 0.0:
dist = normalize_values_in_dict(dist, normalization_factor, inplace=False)
return dist
def predict_proba(self, X):
"""Predicts probabilities of all label of the instance(s).
Parameters
----------
X: numpy.ndarray of shape (n_samples, n_features)
Samples for which we want to predict the labels.
Returns
-------
numpy.array
Predicted the probabilities of all the labels for all instances in X.
"""
r, _ = get_dimensions(X)
predictions = []
for i in range(r):
votes = copy.deepcopy(self.get_votes_for_instance(X[i]))
if votes == {}:
# Tree is empty, all classes equal, default to zero
predictions.append([0])
else:
if sum(votes.values()) != 0:
votes = normalize_values_in_dict(votes, inplace=False)
if self.classes is not None:
y_proba = np.zeros(int(max(self.classes)) + 1)
else:
y_proba = np.zeros(int(max(votes.keys())) + 1)
for key, value in votes.items():
y_proba[int(key)] = value
predictions.append(y_proba)
return np.array(predictions)
def weighted_max(self, X):
""" Get class votes from the rule with highest vote weight.
parameters
----------
X: numpy.ndarray of length equal to the number of features.
Instance attributes.
Returns
-------
dict (class_value, weight)
the class distribution from the rule with highest weight.
"""
highest = 0
final_votes = self.default_rule.get_class_votes(X, self)
for rule in self.rule_set:
if rule.covers_instance(X):
votes = copy.deepcopy(rule.get_class_votes(X, self))
if sum(votes.values()) != 0:
votes = normalize_values_in_dict(votes, inplace=False)
for v in votes.values():
if v >= highest:
highest = v
final_votes = votes
return final_votes
def get_votes_for_instance(self, X):
if self.ensemble is None:
self.ensemble = [
self._init_ensemble_member() for _ in range(self.s)
]
combined_votes = {}
for i in range(self.s):
vote = deepcopy(self.ensemble[i].instance_votes(X))
if vote != {} and sum(vote.values()) > 0:
vote = normalize_values_in_dict(vote, inplace=False)
if self.ensemble.prediction != 0:
accuracy = self.ensemble[i].correct / self.ensemble[
i].prediction
else:
accuracy = 0
if accuracy != 0.0:
for k in vote:
vote[k] = vote[k] * accuracy
# Add values
for k in vote:
try:
combined_votes[k] += vote[k]
except KeyError:
combined_votes[k] = vote[k]
return combined_votes
def predict_proba(self, X):
""" Estimates the probability of each sample in X belonging to each of the class-labels.
Class probabilities are calculated as the mean predicted class probabilities
per base estimator.
Parameters
----------
X: numpy.ndarray of shape (n_samples, n_features)
Samples for which we want to predict the class probabilities.
Returns
-------
numpy.ndarray of shape (n_samples, n_classes)
Predicted class probabilities for all instances in X.
If class labels were specified in a `partial_fit` call, the order of the columns
matches `self.classes`.
If classes were not specified, they are assumed to be 0-indexed.
Class probabilities for a sample shall sum to 1 as long as at least one estimators
has non-zero predictions.
If no estimator can predict probabilities, probabilities of 0 are returned.
"""
if self.ensemble is None:
self._init_ensemble(X)
r, _ = get_dimensions(X)
y_proba = []
for i in range(r):
votes = deepcopy(self.get_votes_for_instance(X[i]))
if votes == {}:
# Estimator is empty, all classes equal, default to zero
if self.classes is not None:
y_proba.append(np.zeros(len(self.classes)))
else:
y_proba.append([0])
else:
if sum(votes.values()) != 0:
votes = normalize_values_in_dict(votes)
if self.classes is not None:
votes_array = np.zeros(int(max(self.classes)) + 1)
else:
votes_array = np.zeros(int(max(votes.keys())) + 1)
for key, value in votes.items():
votes_array[int(key)] = value
y_proba.append(votes_array)
# Set result as np.array
if self.classes is not None:
y_proba = np.asarray(y_proba)
else:
# Fill missing values related to unobserved classes to ensure we get a 2D array
y_proba = np.asarray(
list(itertools.zip_longest(*y_proba, fillvalue=0.0))).T
return y_proba
def get_votes_for_instance(self, X):
if self.ensemble is None:
self.init_ensemble(X)
combined_votes = {}
for i in range(self.n_estimators):
vote = self.ensemble[i].get_votes_for_instance(X)
if vote != {} and sum(vote.values()) > 0:
normalize_values_in_dict(vote)
if not self.disable_weighted_vote:
performance = self.ensemble[i].evaluator.get_accuracy()\
if self.performance_metric == 'acc'\
else self.ensemble[i].evaluator.get_kappa()
if performance != 0.0: # CHECK How to handle negative (kappa) values?
for k in vote:
vote[k] = vote[k] * performance
# Add values
for k in vote:
try:
combined_votes[k] += vote[k]
except KeyError:
combined_votes[k] = vote[k]
return combined_votes
def get_votes_for_instance(self, X):
combined_votes = {}
self.estimators_votes = None
for i in range(len(self.ensemble)):
vote = cp.deepcopy(self.ensemble[i].get_votes_for_instance(X))
if hasattr(self.ensemble[i], 'predict_proba'):
ensemble_class_distribution = self.ensemble[i].predict_proba([X])
if self.estimators_votes is None:
self.estimators_votes = ensemble_class_distribution
else:
self.estimators_votes = np.concatenate(
(self.estimators_votes, ensemble_class_distribution),
axis=1
)
if vote != {} and sum(vote.values()) > 0:
vote = normalize_values_in_dict(vote, inplace=True)
y_proba_dict = None
if self.classes:
y_proba = np.zeros(int(max(self.classes)) + 1)
y_proba_dict = {index: value for index, value in enumerate(y_proba)}
performance = self.ensemble[i].evaluator.accuracy_score()
if performance != 0.0:
for k in vote:
# Multiplying the votes by the performance of each the hoeffding tees in the ensemble
vote[k] = vote[k] * performance
if y_proba_dict:
for key, value in vote.items():
y_proba_dict[float(key)] = value
y_proba_dict = vote
# Add values
for k in vote:
try:
# Combining the result predicted by each classifier for each instance
# combined_votes[k] += vote[k]
combined_votes[k] += y_proba_dict[k]
except KeyError:
# combined_votes[k] = vote[k]
combined_votes[k] = y_proba_dict[k]
return combined_votes
def predict_proba(self, X):
r, _ = get_dimensions(X)
y_proba = []
for i in range(r):
# Calculating the probability of each class using hoeffding trees in the ensemble for the current instance
# (current batch of instances)
votes = cp.deepcopy(self.get_votes_for_instance(X[i]))
if votes == {}:
if self.classes:
y = [0 for i in range(len(self.classes))]
y_proba.append(y)
else:
y_proba.append([0])
# y_proba.append([i for i in range(max(self.classes) +1))])
else:
if sum(votes.values()) != 0:
# Normalizing the votes by dividing each vote from the sum of all the votes
votes = normalize_values_in_dict(votes)
if self.classes is not None:
votes_array = np.zeros(int(max(self.classes)) + 1)
else:
votes_array = np.zeros(int(max(votes.keys())) + 1)
for key, value in votes.items():
try:
votes_array[int(key)] = value
except:
print('this is not ok ')
y_proba.append(votes_array)
if self.classes is not None:
y_proba = np.asarray(y_proba)
else:
y_proba = np.asarray(list(itertools.zip_longest(*y_proba, fillvalue=0.0))).T
# if np.shape(y_proba)[1] == 2:
# import pudb; pudb.set_trace() # XXX BREAKPOINT
# assert 1 == 1
return y_proba
def predict_proba(self, X):
""" Predicts probabilities of all label of the X instance(s)
Parameters
----------
X: numpy.ndarray of shape (n_samples, n_features)
Samples for which we want to predict the labels.
Returns
-------
numpy.array
Predicted the probabilities of all the labels for all instances in X.
"""
r, _ = get_dimensions(X)
predictions = []
for i in range(r):
votes = copy.deepcopy(self.get_votes_for_instance(X[i]))
if votes == {}:
# Tree is empty, all classes equal, default to zero
predictions.append([0])
else:
if sum(votes.values()) != 0:
votes = normalize_values_in_dict(votes, inplace=False)
if self.classes is not None:
y_proba = np.zeros(int(max(self.classes)) + 1)
else:
y_proba = np.zeros(int(max(votes.keys())) + 1)
for key, value in votes.items():
y_proba[int(key)] = value
predictions.append(y_proba)
if len(y_proba) != 2:
assert 1 == 1
# Set result as np.array
if self.classes is not None:
predictions = np.asarray(predictions)
else:
# Fill missing values related to unobserved classes to ensure we get a 2D array
predictions = np.asarray(
list(itertools.zip_longest(*predictions, fillvalue=0.0))).T
return predictions
def predict_one(self, X, *, tree=None):
prediction_option = tree.leaf_prediction
# MC
if prediction_option == tree._MAJORITY_CLASS:
dist = self.stats
# NB
elif prediction_option == tree._NAIVE_BAYES:
dist = do_naive_bayes_prediction(X, self.stats,
self.attribute_observers)
# NBAdaptive (default)
else:
dist = super().predict_one(X, tree=tree)
dist_sum = sum(dist.values()) # sum all values in dictionary
normalization_factor = dist_sum * self.error_estimation * self.error_estimation
if normalization_factor > 0.0:
dist = normalize_values_in_dict(dist,
normalization_factor,
inplace=False)
return dist