def get_distribution(self, instance, class_attribute):
dist = [0.0 for i in range(class_attribute.num_values())]
for i in range(class_attribute.num_values()):
mass = self.class_distribution.get(class_attribute.value(i), None)
if mass is not None:
dist[i] = mass.weight
else:
dist[i] = 1.0
utils.normalize(dist)
return dist
def get_distribution(self, instance, class_attribute):
dist = [0.0 for i in range(class_attribute.num_values())]
for i in range(class_attribute.num_values()):
mass = self.class_distribution.get(class_attribute.value(i), None)
if mass is not None:
dist[i] = mass.weight
else:
dist[i] = 1.0
utils.normalize(dist)
return dist
def distribution_for_instance(self, instance):
"""Return the class probabilities for an instance.
Args:
instance (Instance): The instance to calculate the class probabilites for.
Returns:
list[float]: The class probabilities.
"""
class_attribute = instance.class_attribute()
pred = []
if self._root is not None:
l = self._root.leaf_for_instance(instance, None, None)
actual_node = l.the_node
if actual_node is None:
actual_node = l.parent_node
pred = actual_node.get_distribution(instance, class_attribute)
else:
# All class values equally likely
pred = [1 for i in range(class_attribute.num_values())]
utils.normalize(pred)
return pred
def distribution_for_instance(self, instance):
"""Return the class probabilities for an instance.
Args:
instance (Instance): The instance to calculate the class probabilites for.
Returns:
list[float]: The class probabilities.
"""
class_attribute = instance.class_attribute()
pred = []
if self._root is not None:
l = self._root.leaf_for_instance(instance, None, None)
actual_node = l.the_node
if actual_node is None:
actual_node = l.parent_node
pred = actual_node.get_distribution(instance, class_attribute)
else:
# All class values equally likely
pred = [1 for i in range(class_attribute.num_values())]
utils.normalize(pred)
return pred
def get_scores(self, samples):
return normalize(self.classifier.decision_function(samples).tolist(),
absolute=True)
